Underwater robot sensor fault diagnosis method and system based on machine learning
Through machine learning-based fault diagnosis methods, a fault diagnosis model is built and integrated into the control system of underwater robots, which solves the noise interference problem of depth sensor fault diagnosis in underwater environments, and achieves more efficient and accurate fault detection and positioning , ensuring the smooth completion of robot tasks and equipment safety.
Patent Information
- Application Number
- CN202411291560.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-14
- Publication Date
- 2024-12-24
AI Technical Summary
In underwater environments, the fault diagnosis of depth sensors is greatly disturbed by environmental noise, resulting in low diagnostic accuracy and efficiency, making it difficult to effectively reduce the impact of noise and achieve accurate fault classification.
Using machine learning-based fault diagnosis methods, we collect and preprocess the data of depth sensors, extract fault characteristics, build fault diagnosis models, and integrate them into the control system of underwater robots to perform fault diagnosis and positioning in real time.
It improves the accuracy and efficiency of fault diagnosis, adapts to different underwater environmental conditions and operating modes, and ensures the safe and efficient operation of robot tasks.
Smart Images
Figure CN119178450A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of machine learning, and in particular relates to a method and system for diagnosing underwater robot sensor faults based on machine learning. Background Art
[0002] Underwater robots play a vital role in performing deep-sea exploration missions and participating in the development of marine resources. When these robots perform tasks, depth sensors are one of the core devices, and the accuracy and reliability of their measurement data directly determine the success or failure of the mission. Since deep-sea operations are usually accompanied by high costs and huge risks, once a depth sensor fails, it may lead to the failure of the entire mission and even cause the loss of equipment. Therefore, it is particularly important to be able to diagnose and solve depth sensor failures in a timely and accurate manner.
[0003] Under existing technical conditions, the underwater environment is extremely complex and changeable, and there are various sources of environmental noise, such as water flow noise, biological activity noise, etc. These environmental noises will interfere with the measurement data of the depth sensor to a certain extent, thereby affecting the accuracy of fault diagnosis. In addition, the redundant existence of underwater environmental noise will also increase the burden of fault diagnosis and reduce diagnostic efficiency. Therefore, how to effectively reduce the impact of underwater environmental noise on fault diagnosis and effectively classify the fault diagnosis results of depth sensors has become a key issue that needs to be solved urgently. Summary of the invention
[0004] In order to solve the technical problems existing in the known technology, the present invention provides a method and system for underwater robot sensor fault diagnosis based on machine learning. The present invention uses a fault diagnosis model constructed by machine learning to perform intelligent fault diagnosis on the sensors on the underwater robot.
[0005] The first object of the present invention is to provide an underwater robot sensor fault diagnosis system based on machine learning, which is used to diagnose faults of depth sensors on underwater robots. The method comprises:
[0006] S1, collecting basic data of the depth sensor, wherein the basic data includes normal state data and known fault state data of the underwater robot under different environmental conditions;
[0007] S2, preprocessing the basic data to obtain depth sensor data, and extracting corresponding fault diagnosis feature data according to specific representations of the depth sensor fault;
[0008] S3, marking the corresponding fault state and normal state in the depth sensor data, and determining the corresponding fault type of the fault state to obtain a fault classification sequence table;
[0009] S4, combining the fault classification sequence table and the fault diagnosis feature data, dividing the depth sensor data into a training set, a validation set and a test set, building a fault diagnosis model using a machine learning algorithm, and importing the training set, the validation set and the test set into the fault diagnosis model for training, validation and testing, to obtain a fault diagnosis coefficient to distinguish the depth sensor status;
[0010] S5, integrate the fault diagnosis model into the control system of the underwater robot, collect the real-time data of the depth sensor, use the fault diagnosis model to perform fault diagnosis on the real-time data, and identify and locate the specific fault type of the depth sensor.
[0011] Preferably, S1 comprises:
[0012] S101, clarify the data on the performance change of the depth sensor under different environmental conditions to be obtained, define the environmental condition range for data collection, the environmental condition range includes different water depths, water flow speeds, water temperatures, water salinity, and turbidity; clarify the fault types to be detected, the fault types include sensor offset, data distortion, and response delay;
[0013] S102, collecting basic data of the underwater robot under different environmental conditions, the basic data including historical data and real-time data, and dividing the historical data into normal state data and known fault state data;
[0014] S103, extracting the type, fault occurrence time, condition and manifestation of the known fault status data, and filtering out environmental variable data of the fault status, wherein the environmental variable data includes water depth, water flow velocity and water temperature.
[0015] Preferably, S2 comprises:
[0016] S201, preprocessing the basic data to obtain depth sensor data, wherein the preprocessing includes data cleaning and data smoothing, and standardizing the depth sensor data to a uniform scale range;
[0017] S202, extracting features characterizing sensor states and failure modes from the depth sensor data, and analyzing associated features of failure types to identify failure types that are sensitive to sensor offset, data distortion, and response delay;
[0018] S203, the sensor offset is analyzed by extracting a fixed deviation between the depth reading and the actual depth, the data distortion is analyzed by calculating the data fluctuation range, the standard deviation and the deviation from the normal value, and the response delay is analyzed by analyzing the response time of the sensor reading to the depth change and calculating the delay time.
[0019] Preferably, S3 includes:
[0020] S301, marking the depth sensor data to indicate the normal state and fault state of each data segment;
[0021] S302, for each data segment marked as a fault state, using the extracted fault type association features to analyze the specific fault type of the fault state;
[0022] S303, marking corresponding fault types and fault degrees according to the classified fault status data segments;
[0023] S304, establish a fault classification sequence table, organize the labeled data into the fault classification sequence table in chronological order, and list all data samples and their corresponding states and fault types to ensure that each data point has a corresponding state label and fault degree label.
[0024] Preferably, S4 includes:
[0025] S401, obtaining a fault classification sequence table, wherein each data point has a labeled data state, fault type and fault severity description, and extracting feature data for fault diagnosis, randomly dividing the labeled data into a training set, a validation set and a test set, each set containing samples of various fault types and fault severity, the data state is divided into normal and abnormal, and the fault severity is divided into slight, medium and severe;
[0026] S402, constructing a fault diagnosis model using a machine learning algorithm, training the fault diagnosis model using training set data, optimizing the performance of the model by adjusting model parameters, evaluating the performance of the model using validation set and test set data, and checking the performance of the model on different fault types and degrees;
[0027] S403, based on the fault diagnosis model and the environmental condition range of data collection, obtain a fault diagnosis coefficient, analyze the sensor offset, data distortion, response delay and environmental condition abnormalities of the depth sensor, calculate the standardized deviation of each abnormal state, and distinguish the depth sensor state;
[0028] S404, based on historical data and fault severity data of fault types, statistically analyzing the distribution of different fault types and fault severity, analyzing fault diagnosis coefficients, and classifying fault levels; the fault levels include low fault level, medium fault level, and high fault level;
[0029] S405, setting a corresponding fault warning threshold according to different fault levels, and comparing the current fault diagnosis coefficient with the fault warning threshold to determine the fault level corresponding to the fault type.
[0030] Preferably, S5 includes:
[0031] S501, integrating the fault diagnosis model into the control system of the underwater robot, so that the fault diagnosis model receives data from the depth sensor and returns the diagnosis result;
[0032] S502, the underwater robot obtains real-time data from the sensor and pre-processes the real-time data;
[0033] S503, extracting features related to fault diagnosis from the preprocessed real-time data, and analyzing the real-time data using a fault diagnosis model to implement fault diagnosis of the depth sensor;
[0034] S504, analyzing the fault diagnosis coefficient output by the fault diagnosis model to determine whether the depth sensor has a fault, and identifying the specific fault type and determining the corresponding fault level according to the fault diagnosis coefficient and the output of the fault diagnosis model.
[0035] Preferably, the method further includes S6, issuing a fault diagnosis report and corresponding processing suggestions according to the fault diagnosis result to implement a fault repair or operation adjustment strategy; specifically:
[0036] S601, summarizing and arranging information on the results output by the fault diagnosis model, the fault diagnosis coefficient, the identified fault type, and the determined fault level;
[0037] S602, compiling a fault diagnosis report based on the summarized fault diagnosis information, wherein the fault diagnosis report includes the fault type, fault level, time of the fault occurrence, and impact range information, and records the sensor readings and environmental conditions when the fault occurs, and uses charts and / or images to display the fault data, diagnosis process, and results;
[0038] S603, evaluating the impact of the fault on the current operation, subsequent tasks and the overall system of the underwater robot in combination with the fault diagnosis report, and predicting potential risks based on the fault type and fault degree;
[0039] S604: Based on the fault diagnosis report, formulate specific processing suggestions and repair plans, issue early warning signals, and determine repair measures according to the fault type.
[0040] A second object of the present invention is to provide an underwater robot sensor fault diagnosis system based on machine learning, which is used to diagnose faults of depth sensors on underwater robots. The system comprises:
[0041] A basic data module collects basic data of the depth sensor, wherein the basic data includes normal state data and known fault state data of the underwater robot under different environmental conditions;
[0042] A feature extraction module preprocesses the basic data to obtain depth sensor data, and extracts corresponding fault diagnosis feature data according to the specific representation of the depth sensor fault;
[0043] The fault classification sequence table module marks the corresponding fault state and normal state in the depth sensor data, and determines the corresponding fault type of the fault state to obtain a fault classification sequence table;
[0044] A fault diagnosis model building module, combining the fault classification sequence table and the fault diagnosis feature data, divides the depth sensor data into a training set, a validation set and a test set, uses a machine learning algorithm to build a fault diagnosis model, and imports the training set, the validation set and the test set into the fault diagnosis model for training, validation and testing to obtain a fault diagnosis coefficient to distinguish the depth sensor state;
[0045] The fault diagnosis module integrates the fault diagnosis model into the control system of the underwater robot, collects real-time data of the depth sensor, uses the fault diagnosis model to perform fault diagnosis on the real-time data, and identifies and locates the specific fault type of the depth sensor.
[0046] The third object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned underwater robot sensor fault diagnosis method based on machine learning.
[0047] A fourth object of the present invention is to provide a computer program product, comprising a computer program, which, when executed by a processor, implements the above-mentioned underwater robot sensor fault diagnosis method based on machine learning.
[0048] The advantages and positive effects of the present invention are:
[0049] The present invention uses historical data to train the machine learning algorithm to build a fault diagnosis model, and deeply learns the complex relationship between depth sensor failure and various operating parameters. In this way, the fault diagnosis model can process the data stream from the depth sensor in real time and quickly identify potential signs of failure. This intelligent process includes fault detection, classification and location, and can adapt to different underwater environmental conditions and operating modes, thereby achieving more accurate fault detection. This not only improves the consistency and repeatability of the fault diagnosis process, but also ensures the safety of the robot and the task.
[0050] In addition, the fault diagnosis model of the present invention is not limited to detecting faults of the depth sensor, but can also automatically generate detailed fault diagnosis reports. Once a problem with the depth sensor is detected, the system will immediately generate a report containing the fault type, severity assessment, and recommended treatment measures. Such a report helps to quickly take repair measures, thereby reducing the interruption time of the robot task caused by the fault. In this way, the present invention not only improves the accuracy of fault detection, but also provides valuable feedback and suggestions to maintenance personnel, ensuring the efficient operation of the robot system and the smooth completion of the task. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] Figure 1 is a flow chart of a first preferred embodiment of the present invention;
[0052] Figure 2 is a flow chart of a second preferred embodiment of the present invention;
[0053] Figure 3 Flow chart of step S3 in a preferred embodiment of the present invention;
[0054] Figure 4 Flow chart of step S4 in a preferred embodiment of the present invention. DETAILED DESCRIPTION
[0055] In order to further understand the content, features and effects of the present invention, the following embodiments are listed and described in detail with reference to the accompanying drawings:
[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the technical solutions in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0057] In order to solve the above problems, this application proposes a method and system for underwater robot sensor fault diagnosis based on machine learning technology. Machine learning technology can process large amounts of complex data and extract valuable information and features from them, providing strong support for fault diagnosis. By training and optimizing the algorithm model, the method can identify and classify the fault types of depth sensors, thereby improving the accuracy and efficiency of fault diagnosis. In this way, the present invention can better cope with various challenges that may arise in deep-sea operations, ensuring the smooth progress of tasks and the safe operation of equipment.
[0058] First embodiment:
[0059] See also Figure 1A method for underwater robot sensor fault diagnosis based on machine learning is used to diagnose faults of deep-water sensors on underwater robots. The method comprises:
[0060] S1, collecting basic data of the depth sensor, wherein the basic data includes normal state data and known fault state data of the underwater robot under different environmental conditions;
[0061] The integrity and richness of basic data is one of the important factors that determine the quality of the fault diagnosis model. In order to ensure the integrity and richness of data, this embodiment adopts the following technical solutions:
[0062] S101, clearly define the data that needs to be obtained on the performance changes of the underwater robot's depth sensor under different environmental conditions, define the environmental conditions for data collection, including different water depths, water flow speeds, water temperatures, water salinity, and turbidity, and clearly define the types of faults that need to be detected, which are divided into sensor offset, data distortion, and response delay. The specific manifestation of sensor offset is that there is a fixed deviation between the sensor reading and the actual depth. The specific manifestation of data distortion is that the sensor reading fluctuates greatly and cannot accurately reflect the actual depth, which may be caused by electromagnetic interference, mechanical vibration, etc. The specific manifestation of response delay is that the sensor updates data too slowly after the depth changes, resulting in poor real-time performance;
[0063] S102, collecting basic data of the underwater robot under different environmental conditions, where the basic data is mainly divided into historical data and real-time data, and the historical data is divided into normal state data and known fault state data;
[0064] S103, for the known fault status data obtained, extract its type, fault occurrence time, conditions and manifestations, and filter out environmental variable data of the fault status. The environmental variable data includes water depth, water flow velocity, and water temperature, which is helpful for subsequent analysis of the impact of environmental factors on sensor performance.
[0065] S2, preprocessing the basic data to obtain depth sensor data, and extracting corresponding fault diagnosis feature data according to the specific representation of the sensor fault; specifically including:
[0066] S201, preprocessing the basic data, including data cleaning and data smoothing, and standardizing the depth sensor data to a uniform scale range to ensure that the values of different features are comparable, wherein data cleaning is used to remove or correct noise, outliers, missing values, etc. in the original data to improve data quality, and data smoothing is used to apply digital filtering technology to smooth the sensor data to reduce the impact of noise on subsequent analysis;
[0067] S202, extracting features characterizing sensor states and fault modes from the preprocessed depth sensor data, analyzing associated features of fault types, and identifying fault types that are sensitive to sensor offset, data distortion, and response delay;
[0068] S203, the sensor offset is analyzed by extracting the fixed deviation between the depth reading and the actual depth, the data distortion is analyzed by calculating the data fluctuation range, the standard deviation and the deviation from the normal value, and the response delay is analyzed by analyzing the response time of the sensor reading to the depth change and calculating the delay time.
[0069] S3, marking the corresponding fault state and normal state in the depth sensor data, and determining the corresponding fault type of the fault state to obtain a fault classification sequence table; please refer to Figure 3 , including:
[0070] S301, marking the pre-processed depth sensor data, indicating the normal state and fault state of each data segment;
[0071] S302, for each data segment marked as a fault state, further analyzing the specific fault type of the fault state using the extracted fault type association features;
[0072] S303, marking corresponding fault types and fault degrees according to the classified fault status data segments;
[0073] S304, establish a fault classification sequence table, organize the labeled data into the fault classification sequence table in chronological order, and list all data samples and their corresponding states and fault types to ensure that each data point has a corresponding state label and fault degree label.
[0074] S4, combining the fault classification sequence table and the fault diagnosis feature data, dividing the depth sensor data into a training set, a validation set and a test set, using a machine learning algorithm to build a fault diagnosis model, and importing the training set, the validation set and the test set into the fault diagnosis model for training, validation and testing to obtain a fault diagnosis coefficient to distinguish the sensor status; please refer to Figure 4 , including:
[0075] S401, using a fault classification sequence table, where each data point has an annotated data state, fault type and fault severity description, and extracting feature data for fault diagnosis, randomly dividing the annotated data into three subsets: training set, validation set and test set. The data set division should ensure that each set contains samples of various fault types and fault severity to ensure the performance of the model on diversified data. The data state is divided into normal and abnormal, the fault type is divided into sensor offset, data distortion, response delay, and the fault severity is divided into mild, moderate and severe.
[0076] S402, constructing a fault diagnosis model according to a machine learning algorithm, training the fault diagnosis model using training set data, optimizing the performance of the model by adjusting model parameters, evaluating the performance of the model using validation set and test set data, and checking the performance of the model on different fault types and degrees;
[0077] S403, based on the fault diagnosis model and in combination with the environmental condition range of data collection, a fault diagnosis coefficient is obtained, a sensor offset, data distortion, response delay and abnormal environmental conditions of the depth sensor are comprehensively analyzed, a standardized deviation of each abnormal state is calculated, and a state of the depth sensor is distinguished;
[0078] S404, based on historical data and fault severity data of fault types, statistically analyzing the distribution of different fault types and severity, analyzing fault diagnosis coefficients, and classifying fault levels into low fault level, medium fault level, and high fault level;
[0079] S405, setting a corresponding fault warning threshold according to the different fault levels, and comparing the current fault diagnosis coefficient with the threshold to determine the fault level corresponding to the fault type.
[0080] Further: The calculation formula of the fault diagnosis coefficient is:
[0081]
[0082] Among them, AD is the fault diagnosis coefficient, Δd i is the depth reading deviation of the i-th data point, indicating the difference between the measured value and the standard value, μΔd is the mean of the depth reading deviation, indicating the average value of all deviation values Δd i Find the average, σΔd is the standard deviation of the depth reading deviation, which means the deviation value Δd i The discrete degree of the data is quantified, N is the total number of data points, which refers to the number of samples in the data set, r j is the response delay of the jth data point, which represents the difference between the measured sensor response time and the theoretical response time, σ r is the standard deviation of the response delay, representing the standard deviation of the response delay r j Find the standard deviation, M is the number of response delay data points, c k is the value of the kth environmental condition data point, obtained by measuring the environmental parameters of water temperature, water flow velocity, salinity, and turbidity, μ c and σ c is the mean and standard deviation of the environmental conditions, indicating that for all environmental condition values c k Find the mean and standard deviation, where L is the number of environmental condition data points.
[0083] A further improvement is that: a plurality of the fault levels correspond to a plurality of the fault warning thresholds, wherein the fault warning thresholds include an upper threshold and a lower threshold;
[0084] The multiple fault levels and the multiple fault warning thresholds satisfy the following relationship:
[0085] Low fault level AD≤AD T1 ; For minor impacts, set a lower warning threshold to alert operators;
[0086] Medium fault level AD T1 <AD≤AD T2 ; Moderate impact, set a medium warning threshold and recommend preliminary inspection;
[0087] High fault level AD>AD T2 ; Severe impact, set high warning thresholds, and immediately conduct troubleshooting and repair;
[0088] Among them, AD is the fault diagnosis coefficient, AD T1 is the lower threshold corresponding to the medium fault level and the upper threshold corresponding to the low fault level, AD T2 It is the lower threshold corresponding to the high fault level and the upper threshold corresponding to the medium fault level.
[0089] S5, integrate the fault diagnosis model into the control system of the underwater robot, collect the real-time data of the sensor, use the fault diagnosis model to perform fault diagnosis on the real-time data, and identify and locate the specific fault type of the sensor. Specifically include:
[0090] S501, integrating the trained fault diagnosis model into the control system of the underwater robot, so that the fault diagnosis model receives data from the depth sensor and returns the diagnosis result;
[0091] S502, the underwater robot collects environmental data in real time through a depth sensor, and pre-processes the real-time collected data to ensure data quality and consistency;
[0092] S503, extracting key features related to fault diagnosis from the preprocessed real-time data, and analyzing the real-time data using a fault diagnosis model to implement fault diagnosis of the depth sensor;
[0093] S504, analyzing the fault diagnosis coefficient output by the fault diagnosis model to determine whether the sensor has a fault, and identifying the specific fault type and the corresponding fault level according to the fault diagnosis coefficient and the output of the fault diagnosis model.
[0094] Second embodiment:
[0095] See also Figure 2, a method for underwater robot sensor fault diagnosis based on machine learning, based on the above first preferred embodiment, further comprising S6, issuing a fault diagnosis report and corresponding processing suggestions according to the fault diagnosis result to implement fault repair or operation adjustment strategy; specifically:
[0096] S601, summarizing and arranging information on the results output by the fault diagnosis model, the fault diagnosis coefficient, the identified fault type, and the determined fault level;
[0097] S602, compile a fault diagnosis report based on the summarized fault diagnosis information, including the fault type, fault level, time of fault occurrence, and impact range information, and record the sensor readings and environmental conditions when the fault occurred, and use auxiliary means such as charts and images to intuitively display the fault data, diagnosis process and results;
[0098] S603, evaluating the impact of the fault on the underwater robot's current operation, subsequent tasks, and the overall system in combination with the fault diagnosis report, and predicting potential risks based on the type and degree of the fault;
[0099] S604: Based on the fault diagnosis report, formulate specific processing suggestions and repair plans, and send early warning signals to relevant personnel. Determine repair measures based on the type of fault, such as replacing sensor components, recalibrating or adjusting sensor settings, etc. If the fault affects the robot's operating ability or task completion ability, provide suggestions for adjusting the operation plan or task allocation.
[0100] Third embodiment:
[0101] A machine learning-based underwater robot sensor fault diagnosis system is used to diagnose faults of depth sensors on underwater robots. The system comprises:
[0102] A basic data module collects basic data of the depth sensor, wherein the basic data includes normal state data and known fault state data of the underwater robot under different environmental conditions; the working process of the basic data module is the same as that of the first embodiment;
[0103] A feature extraction module preprocesses the basic data to obtain depth sensor data, and extracts corresponding fault diagnosis feature data according to the specific representation of the sensor fault; the feature extraction processing of the feature extraction module is the same as that of the first embodiment;
[0104] The fault classification sequence table module marks the corresponding fault state and normal state in the depth sensor data, and determines the corresponding fault type of the fault state to obtain a fault classification sequence table; the working process of the fault classification sequence table module is the same as that of the first embodiment;
[0105] A fault diagnosis model building module, combining the fault classification sequence table and the fault diagnosis feature data, divides the depth sensor data into a training set, a validation set and a test set, uses a machine learning algorithm to build a fault diagnosis model, and imports the training set, the validation set and the test set into the fault diagnosis model for training, validation and testing to obtain a fault diagnosis coefficient to distinguish the sensor state; the working process of the fault diagnosis model building module is the same as that of the first embodiment;
[0106] The fault diagnosis module integrates the fault diagnosis model into the control system of the underwater robot, collects the real-time data of the sensor, uses the fault diagnosis model to perform fault diagnosis on the real-time data, and identifies and locates the specific fault type of the sensor. The working process of the fault diagnosis module is the same as that of the first embodiment;
[0107] Fourth embodiment: A machine learning-based underwater robot sensor fault diagnosis system, used for fault diagnosis of a depth sensor on an underwater robot, the system comprising: on the basis of the above-mentioned first preferred embodiment, further comprising an output module, which issues a fault diagnosis report and corresponding processing suggestions according to the fault diagnosis result, so as to implement a fault repair or operation adjustment strategy; specifically:
[0108] Summarize and organize the results of the fault diagnosis model output, the fault diagnosis coefficient, the identified fault type, and the determined fault level information;
[0109] Prepare a fault diagnosis report based on the summarized fault diagnosis information, including the fault type, fault level, time of fault occurrence, and impact range information, and record the sensor readings and environmental conditions when the fault occurred. Use charts, images and other auxiliary means to intuitively display the fault data, diagnosis process and results;
[0110] Combined with the fault diagnosis report, evaluate the impact of the fault on the underwater robot's current operation, subsequent tasks and the overall system, and predict potential risks based on the type and degree of the fault;
[0111] Based on the fault diagnosis report, formulate specific processing suggestions and repair plans, and send early warning signals to relevant personnel. Determine repair measures based on the type of fault, such as replacing sensor components, recalibrating or adjusting sensor settings, etc. If the fault affects the robot's operating ability or task completion ability, provide suggestions for adjusting the operation plan or task allocation.
[0112] The fault diagnosis model in this application can select the OS-ELM model, which is a neural network model that can be used for fault diagnosis of depth sensors.
[0113] Fifth embodiment: A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned underwater robot sensor fault diagnosis method based on machine learning.
[0114] Sixth embodiment: A computer program product, comprising a computer program, which, when executed by a processor, implements the above-mentioned underwater robot sensor fault diagnosis method based on machine learning.
[0115] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When the use is implemented in whole or in part in the form of 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, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL) or wireless (e.g., infrared, wireless, microwave, etc.) mode) to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk SolidState Disk (SSD)), etc.
[0116] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention are within the scope of the technical solution of the present invention.
Claims
1. A method for underwater robot sensor fault diagnosis based on machine learning, characterized in that: For performing fault diagnosis on a depth sensor on an underwater robot, the method comprises: S1, collecting basic data of the depth sensor, wherein the basic data includes normal state data and known fault state data of the underwater robot under different environmental conditions; S2, preprocessing the basic data to obtain depth sensor data, and extracting corresponding fault diagnosis feature data according to specific representations of the depth sensor fault; S3, marking the corresponding fault state and normal state in the depth sensor data, and determining the corresponding fault type of the fault state to obtain a fault classification sequence table; S4, combining the fault classification sequence table and the fault diagnosis feature data, dividing the depth sensor data into a training set, a validation set and a test set, building a fault diagnosis model using a machine learning algorithm, and importing the training set, the validation set and the test set into the fault diagnosis model for training, validation and testing, to obtain a fault diagnosis coefficient to distinguish the depth sensor status; S5, integrate the fault diagnosis model into the control system of the underwater robot, collect the real-time data of the depth sensor, use the fault diagnosis model to perform fault diagnosis on the real-time data, and identify and locate the specific fault type of the depth sensor.
2. The underwater robot sensor fault diagnosis method based on machine learning according to claim 1 is characterized in that: S1 includes: S101, clarify the data on the performance change of the depth sensor under different environmental conditions to be obtained, define the environmental condition range for data collection, the environmental condition range includes different water depths, water flow speeds, water temperatures, water salinity, and turbidity; clarify the fault types to be detected, the fault types include sensor offset, data distortion, and response delay; S102, collecting basic data of the underwater robot under different environmental conditions, the basic data including historical data and real-time data, and dividing the historical data into normal state data and known fault state data; S103, extracting the type, fault occurrence time, condition and manifestation of the known fault status data, and filtering out environmental variable data of the fault status, wherein the environmental variable data includes water depth, water flow velocity and water temperature.
3. The underwater robot sensor fault diagnosis method based on machine learning according to claim 2 is characterized in that: S2 includes: S201, preprocessing the basic data to obtain depth sensor data, wherein the preprocessing includes data cleaning and data smoothing, and standardizing the depth sensor data to a uniform scale range; S202, extracting features characterizing sensor states and failure modes from the depth sensor data, and analyzing associated features of failure types to identify failure types that are sensitive to sensor offset, data distortion, and response delay; S203, the sensor offset is analyzed by extracting a fixed deviation between the depth reading and the actual depth, the data distortion is analyzed by calculating the data fluctuation range, the standard deviation and the deviation from the normal value, and the response delay is analyzed by analyzing the response time of the sensor reading to the depth change and calculating the delay time.
4. The underwater robot sensor fault diagnosis method based on machine learning according to claim 3 is characterized in that S3 include: S301, marking the depth sensor data to indicate the normal state and fault state of each data segment; S302, for each data segment marked as a fault state, using the extracted fault type association features to analyze the specific fault type of the fault state; S303, marking corresponding fault types and fault degrees according to the classified fault status data segments; S304, establish a fault classification sequence table, organize the labeled data into the fault classification sequence table in chronological order, and list all data samples and their corresponding states and fault types to ensure that each data point has a corresponding state label and fault degree label.
5. The underwater robot sensor fault diagnosis method based on machine learning according to claim 4 is characterized in that S4 include: S401, obtaining a fault classification sequence table, wherein each data point has a labeled data state, fault type and fault severity description, and extracting feature data for fault diagnosis, randomly dividing the labeled data into a training set, a validation set and a test set, each set containing samples of various fault types and fault severity, the data state is divided into normal and abnormal, and the fault severity is divided into slight, medium and severe; S402, constructing a fault diagnosis model using a machine learning algorithm, training the fault diagnosis model using training set data, optimizing the performance of the model by adjusting model parameters, evaluating the performance of the model using validation set and test set data, and checking the performance of the model on different fault types and degrees; S403, based on the fault diagnosis model and the environmental condition range of data collection, obtain a fault diagnosis coefficient, analyze the sensor offset, data distortion, response delay and environmental condition abnormalities of the depth sensor, calculate the standardized deviation of each abnormal state, and distinguish the depth sensor state; S404, based on historical data and fault severity data of fault types, statistically analyzing the distribution of different fault types and fault severity, analyzing fault diagnosis coefficients, and classifying fault levels; the fault levels include low fault level, medium fault level, and high fault level; S405, setting a corresponding fault warning threshold according to different fault levels, and comparing the current fault diagnosis coefficient with the fault warning threshold to determine the fault level corresponding to the fault type.
6. The underwater robot sensor fault diagnosis method based on machine learning according to claim 5 is characterized in that S5 include: S501, integrating the fault diagnosis model into the control system of the underwater robot, so that the fault diagnosis model receives data from the depth sensor and returns the diagnosis result; S502, the underwater robot obtains real-time data from the sensor and pre-processes the real-time data; S503, extracting features related to fault diagnosis from the preprocessed real-time data, and analyzing the real-time data using a fault diagnosis model to implement fault diagnosis of the depth sensor; S504, analyzing the fault diagnosis coefficient output by the fault diagnosis model to determine whether the depth sensor has a fault, and identifying the specific fault type and determining the corresponding fault level according to the fault diagnosis coefficient and the output of the fault diagnosis model.
7. The underwater robot sensor fault diagnosis method based on machine learning according to claim 6 is characterized in that: It also includes S6, which issues a fault diagnosis report and corresponding processing suggestions according to the fault diagnosis result to implement fault repair or operation adjustment strategy; Specifically: S601, summarizing and arranging information on the results output by the fault diagnosis model, the fault diagnosis coefficient, the identified fault type, and the determined fault level; S602, compiling a fault diagnosis report based on the summarized fault diagnosis information, wherein the fault diagnosis report includes the fault type, fault level, time of the fault occurrence, and impact range information, and records the sensor readings and environmental conditions when the fault occurs, and uses charts and / or images to display the fault data, diagnosis process, and results; S603, evaluating the impact of the fault on the current operation, subsequent tasks and the overall system of the underwater robot in combination with the fault diagnosis report, and predicting potential risks based on the fault type and fault degree; S604: Based on the fault diagnosis report, formulate specific processing suggestions and repair plans, issue early warning signals, and determine repair measures according to the fault type.
8. A sensor fault diagnosis system for underwater robots based on machine learning, characterized in that: For diagnosing faults of a depth sensor on an underwater robot, the system comprises: A basic data module collects basic data of the depth sensor, wherein the basic data includes normal state data and known fault state data of the underwater robot under different environmental conditions; A feature extraction module preprocesses the basic data to obtain depth sensor data, and extracts corresponding fault diagnosis feature data according to the specific representation of the depth sensor fault; The fault classification sequence table module marks the corresponding fault state and normal state in the depth sensor data, and determines the corresponding fault type of the fault state to obtain a fault classification sequence table; A fault diagnosis model building module, combining the fault classification sequence table and the fault diagnosis feature data, divides the depth sensor data into a training set, a validation set and a test set, uses a machine learning algorithm to build a fault diagnosis model, and imports the training set, the validation set and the test set into the fault diagnosis model for training, validation and testing to obtain a fault diagnosis coefficient to distinguish the depth sensor state; The fault diagnosis module integrates the fault diagnosis model into the control system of the underwater robot, collects real-time data of the depth sensor, uses the fault diagnosis model to perform fault diagnosis on the real-time data, and identifies and locates the specific fault type of the depth sensor.
9. A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the underwater robot sensor fault diagnosis method based on machine learning as described in any one of claims 1 to 7.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the underwater robot sensor fault diagnosis method based on machine learning described in any one of claims 1 to 7 is implemented.