CN114662327A - Auxiliary method for leakage monitoring in AUV (autonomous Underwater vehicle) cabin - Google Patents

Auxiliary method for leakage monitoring in AUV (autonomous Underwater vehicle) cabin Download PDF

Info

Publication number
CN114662327A
CN114662327A CN202210320437.8A CN202210320437A CN114662327A CN 114662327 A CN114662327 A CN 114662327A CN 202210320437 A CN202210320437 A CN 202210320437A CN 114662327 A CN114662327 A CN 114662327A
Authority
CN
China
Prior art keywords
cabin
deformation
leakage
data
auv
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210320437.8A
Other languages
Chinese (zh)
Inventor
高爽
钟大林
严天宏
沈钺
何波
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Qingdao Pengpai Ocean Exploration Technology Co ltd
Original Assignee
Qingdao Pengpai Ocean Exploration Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Qingdao Pengpai Ocean Exploration Technology Co ltd filed Critical Qingdao Pengpai Ocean Exploration Technology Co ltd
Priority to CN202210320437.8A priority Critical patent/CN114662327A/en
Publication of CN114662327A publication Critical patent/CN114662327A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Algebra (AREA)
  • Computing Systems (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • Examining Or Testing Airtightness (AREA)

Abstract

The invention discloses an auxiliary method for monitoring leakage in an AUV (autonomous underwater vehicle) cabin, which is characterized in that in the running process of the AUV, aiming at the false alarm phenomenon of leakage monitoring in the cabin, the influence of false alarm is eliminated by using a method based on data fusion of an internal sensor, running states such as temperature, humidity, air pressure, leakage, depth, cabin deformation, motor rotating speed, voltage, current and the like are monitored by using a sensor carried by the AUV, and the most main state variable related to leakage is extracted by using correlation analysis; and processing deformation data by applying a Kalman filtering mathematical model to two important parameters of cabin deformation and depth to obtain a deformation optimal value, and finally assisting to monitor the leakage in the underwater vehicle cabin by using an auto-encoder method based on the fusion of various sensor numbers. Additional sensors are not introduced, leakage auxiliary monitoring is achieved by using basic cabin state monitoring data, monitoring results are accurate and reliable, and system reliability is improved.

Description

Auxiliary method for leakage monitoring in AUV (autonomous Underwater vehicle) cabin
Technical Field
The invention belongs to the technical field of leakage monitoring in an underwater vehicle cabin, and particularly relates to an auxiliary method for leakage monitoring in an AUV cabin.
Background
Autonomous underwater vehicles are now gaining more and more widespread acceptance in the fields of national defense, marine geology, industry and the like. However, as the application field is expanded, the AUV needs to monitor and alarm whether leakage occurs during task execution. At present, leakage detection of the AUV usually adopts a short-circuit terminal form. Namely, a short circuit detection circuit made of an FPC flexible circuit board is attached to a position prone to leakage. Short circuit detection can be triggered when water exists on the FPC, and then leakage alarm information is generated, but the AUV is an extremely complex electromechanical system, and the AUV is internally provided with a large number of electronic elements and mechanical parts, so that the structure is broken, metal dust, electronic components are broken, and the like, which can cause the AUV to generate leakage false alarm.
Once the internal leak detection module gives an alarm, whether it is a true leak alarm or a false alarm, the equipment is almost taken apart for maintenance. If an alarm occurs during the operation of the task, the equipment can only float up emergently. This will significantly affect the stability and safety of the AUV, and also increase the workload.
In the prior art, there is almost no relevant solution for false alarm, so in order to ensure that the AUV can complete the expected task safely, efficiently and reliably, it is necessary to perform auxiliary judgment on the leakage alarm module of the AUV during the task, so as to reduce the influence of the false alarm on the equipment safety and the task process.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides an auxiliary method for leakage monitoring in an AUV (autonomous underwater vehicle) cabin, aiming at reducing the influence of false alarm on equipment safety and task progress.
The invention is realized by adopting the following technical scheme: an auxiliary method for leakage monitoring within an AUV capsule, comprising the steps of:
step A, analyzing the correlation between each variable and leakage based on AUV (autonomous underwater vehicle) cabin data to determine a key factor most correlated to cabin leakage;
the variables comprise air pressure, depth, main cabin temperature, main cabin humidity, motor temperature, cabin body deformation, system battery voltage, system battery current, power battery voltage and power battery current, and the determined key factors comprise the depth, the air pressure, the cabin body deformation and the main cabin temperature;
b, cabin deformation optimal estimation based on Kalman filtering;
taking cabin deformation data values actually measured by a plurality of deformation sensors on the AUV cabin as observed values, taking deformation values estimated according to a numerical computation model by using depth values as predicted values, and estimating the optimal deformation of the cabin by combining a Kalman filtering equation based on the observed values and the predicted values;
step C, establishing a leakage self-encoder through related data to carry out leakage abnormity monitoring;
(1) establishing a data set by using historical cabin data of the AUV, wherein the data set comprises six variables of air pressure, main cabin temperature, depth, optimal deformation of a cabin body, initial temperature and initial air pressure, and determining a training set, a verification set and a test set;
(2) constructing and training a leakage self-encoder, taking six variables of test concentration depth, main cabin temperature, air pressure, cabin deformation optimal value, initial cabin temperature and initial air pressure as input, and outputting corresponding 6 reconstruction values after being processed by the leakage self-encoder; the difference value between the reconstruction value and the input is a reconstruction error, and the maximum value of the sum of the absolute values of the reconstruction errors of 6 variables in the test set is used as a judgment threshold;
(3) when the AUV has a leakage sensor to report leakage alarm, inputting the current air pressure, the temperature and the depth of the main cabin, the optimal deformation of the cabin, the initial temperature and the initial air pressure data into a trained leakage self-encoder to obtain the sum of the absolute values of the reconstruction errors of 6 variables at the current moment, recording the sum as a judgment parameter, when the judgment parameter is greater than a judgment threshold value, considering that the leakage is abnormal, otherwise, judging that the leakage is misjudged.
Further, in the step a, when determining the key factor, determining a key parameter related to the air pressure through correlation analysis:
(1) the correlation of the air pressure with other variables was calculated using spearman correlation coefficients:
Figure BDA0003570314930000021
where Ps is a correlation calculated from the spearman correlation coefficient, x represents the level of the air pressure, y represents the level of any one of the other nine variables, i represents the ith variable data, n represents the total number of the variable data,
Figure BDA0003570314930000022
represents the rank average of x and y;
(2) performing hypothesis test on the correlation to provide an original hypothesis H0For Ps 0, assume H1Ps is not equal to 0; calculating statistic z obeying standard normal distribution by taking original hypothesis establishment as background*；
Figure BDA0003570314930000023
Calculating cumulative distribution function of standard normal distribution, and calculating cumulative distribution function value z*And (3) substituting a certain corresponding determined point m into the following formula to obtain a check value p:
p＝2(1-m)
if p is>0.05, the original hypothesis cannot be rejected at a confidence level of 95%, i.e. the original hypothesis H is satisfied0No significant difference exists, and no correlation exists among data; otherwise, reject the original hypothesis at 95% confidence level, i.e., satisfy hypothesis H1And the data have correlation.
Further, the step B is specifically realized by the following steps:
(1) firstly, AUV cabin deformation data is used as observed quantity, a deformation value estimated by using a depth value according to a numerical calculation model y, namely Lx + B, is used as a predicted value, Kalman filtering is used for correcting the predicted value in real time, and iterative calculation is as follows:
Figure BDA0003570314930000031
Zk＝LDk+B+Vk
Figure BDA0003570314930000032
Figure BDA0003570314930000033
Figure BDA0003570314930000034
Figure BDA0003570314930000035
wherein S iskIs an observed value of deformation of the cabin body at the moment k, Wk-1And VkRepresenting the Gaussian noise in the observation and prediction processes, respectively, and the corresponding covariances are Q and R, ZkThe predicted value of the deformation at the time k, D, is obtained from the modelkIs a depth measurement, L and B are numerical calculation model parameters; pk-1Is that
Figure BDA0003570314930000036
Of the covariance matrix, PkIs that
Figure BDA0003570314930000037
The covariance matrix of (a) is determined,
Figure BDA0003570314930000038
is that
Figure BDA0003570314930000039
Covariance matrix of, KkIs the gain of the kalman gain (in),
Figure BDA00035703149300000310
is the optimal deformation quantity at the last moment,
Figure BDA00035703149300000311
is the optimal deformation quantity at the current moment,
Figure BDA00035703149300000312
the predicted value of the k moment is obtained according to the optimal value of the previous moment; wherein
Figure BDA00035703149300000313
Is set to 0, will
Figure BDA00035703149300000314
Figure BDA00035703149300000315
Pk，Pk-1Setting the initial value of the equal covariance matrix as a unit matrix;
(2) obtaining a filtering value through the processing of the step (1), carrying out cyclic iteration on the data according to a time sequence based on the current filtering value, and solving the optimal deformation quantity of the deformation sensors at different positions at the current moment;
(3) after the deformation data collected by the deformation sensors at different positions are filtered, the average deformation data is taken as the final optimal deformation of the cabin body, and the overall deformation degree of the AUV cabin body is measured.
Further, in the step C, the leakage self-encoder includes an input layer, an encoder, a decoder, and an output layer, the encoder and the decoder are structurally symmetric and include two hidden layers, respectively, the two hidden layers MLP structure of the encoder encode and compress the input six feature data, extract the relevant features of four dimensions of output depth, optimal deformation of cabin, temperature of main cabin, and air pressure, and these 4 feature quantities are used as the input quantity of the decoder, and finally, the output layer outputs 6 output features, that is, the reconstruction quantity
Compared with the prior art, the invention has the advantages and positive effects that:
in the scheme, in the AUV operation process, Kalman filtering is utilized to combine cabin deformation data of different positions of an aircraft with a theoretical model of cabin deformation to obtain an optimal value of cabin deformation, and the optimal value is used for eliminating the influence of position interference on cabin deformation; reconstructing variables such as temperature, air pressure, deformation quantity of the cabin body, depth, initial temperature and initial air pressure by using a leakage self-encoder, and subtracting a reconstructed value from a measured value to generate a residual error as a judgment standard of abnormal monitoring;
an additional sensor is not introduced, the leakage auxiliary monitoring is realized by using the basic cabin state monitoring data, the monitoring result is accurate and reliable, and the system reliability is improved; and the whole method is based on the causal relationship of the leakage, the selected variables are strongly related to the leakage, and the monitoring of the abnormal state is realized by using an unsupervised method.
Drawings
FIG. 1 is a schematic flow chart of an auxiliary method according to an embodiment of the present invention;
FIG. 2 is a block diagram of a self-encoding trainer in accordance with an embodiment of the invention;
FIG. 3 is a schematic diagram of a leaky self-encoder according to an embodiment of the invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention more clearly understood, the present invention will be further described with reference to the accompanying drawings and examples. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those described herein, and thus, the present invention is not limited to the specific embodiments disclosed below.
Aiming at the false alarm phenomenon of leakage monitoring in an AUV (autonomous Underwater vehicle) cabin in the running process, the invention provides an auxiliary method for leakage monitoring in the AUV cabin, which has the following basic principle: firstly, extracting key factors related to leakage by using data in an AUV (autonomous Underwater vehicle) cabin; then, obtaining the optimal cabin deformation by using Kalman filtering, and eliminating the influence of position interference on the cabin deformation; and after the original leakage sensor gives an alarm, performing reconstruction detection on the determined key factors, the optimal deformation of the cabin body and other parameters by using a leakage self-encoder, performing difference on a reconstruction value and an input value, and using the generated residual error as a judgment standard of abnormal monitoring to further determine whether the cabin body is leaked.
Specifically, as shown in fig. 1, the auxiliary method for monitoring leakage in the AUV cabin includes the following steps:
step A, analyzing the correlation between each variable and leakage based on AUV (autonomous underwater vehicle) cabin data to determine a key factor most correlated to cabin leakage;
firstly, a large amount of AUV (autonomous underwater vehicle) cabin internal data are utilized to analyze the correlation between each variable and leakage, the cabin leakage and the cabin internal air pressure change have a direct relation under normal conditions, and if abnormal cabin internal air pressure change can be detected, whether the cabin body leaks or not can be indirectly determined, namely, key factors related to the leakage can be determined by determining key parameters related to the air pressure. Generally, monitoring parameters of the state system in the AUV cabin include ten variables, such as air pressure, depth, main cabin temperature \ humidity, motor temperature, cabin deformation, system battery voltage, current, and power battery voltage, current, etc. However, some of these variables are not related to the air pressure, so in order to determine the several variables most related to the air pressure, the present embodiment determines the sensor variable related to the air pressure by using a correlation analysis, which is as follows:
monitoring parameters (ten variables such as air pressure, depth, main cabin temperature and humidity, motor temperature, cabin deformation, system battery and power battery voltage and current) of an AUV cabin internal state system, wherein the deformation is fixed-point sampling monitoring. Experience shows that the pressure applied to different positions of the cabin body is different, the closer the cabin body is to the supporting structure, the higher the compressive strength is, and therefore, the data recorded by the strain gauge are different. The deformation data used in the following all need to reflect the whole deformation degree of the AUV, so that the deformation data acquired by the deformation sensors at different positions are averaged, and the average value of the deformation data is used as a standard to measure the whole deformation degree of the AUV. In this embodiment, the correlations between the air pressure and other variables are calculated by using the spearman correlation coefficient which does not strictly require data conditions, specifically as follows:
(1) and calculating the correlation between the air pressure and the other variables by using the spearman correlation coefficient, wherein the calculation formula is shown as formula 1:
Figure BDA0003570314930000051
where Ps is a correlation calculated from the spearman correlation coefficient, x represents the level of the air pressure, and y represents the level of any one of the other nine variables. The variable grade is that the variable values in a period of time are arranged from low to high, the lower grade of the value is smaller, the higher grade of the value is larger, i represents the ith variable data, n represents the total number of the data,
Figure BDA0003570314930000052
the rank average of x and y is indicated.
(2) After calculating the correlation between the air pressure and other variables, a hypothesis test is performed on the correlation. Propose the original hypothesis H0For Ps to be 0, assume H1Ps ≠ 0. Calculating statistic z obeying standard normal distribution by taking the establishment of original hypothesis as background*Such as
Formula 1-2;
Figure BDA0003570314930000053
a cumulative distribution function of a normal distribution (the cumulative distribution function is an integral of a probability density function which is a function describing the probability of the output value of the random variable in the vicinity of a certain value-taking point) is calculated, and a cumulative distribution function value z is calculated*And substituting a certain corresponding determined point m to obtain a test value p, as shown in the formula 1-3.
p＝2(1-m) (1-3)
If p is>0.05, then 95%The original hypothesis cannot be rejected at confidence level, i.e. the original hypothesis H is satisfied0(ρS0), no significant difference, no correlation between data; if p is<0.05, the original hypothesis is rejected at a confidence level of 95%, i.e. hypothesis H is satisfied1(ρSNot equal to 0), has significant difference, and has certain correlation among data; the specific correlation analysis results are shown in table 1:
TABLE 1 correlation coefficient of barometric pressure with other variables
Figure BDA0003570314930000054
Figure BDA0003570314930000061
As can be seen from table 1, the p value of the spearman correlation coefficient of the motor rotation speed is 0.7175 and is greater than 0.05, and therefore the original assumption cannot be rejected, indicating that there is no correlation. The p values of other parameters are less than 0.05, so the original hypothesis can be rejected, and certain correlation exists. After the existence of the correlation is determined, the variables are classified according to the degree of correlation, and the existence of the strong correlation with the air pressure can be obtained: depth, deformation, main compartment temperature, and motor temperature. Because the AUV motor and the cabin body are designed independently and separately, the air pressure change in the AUV cabin is not directly influenced, the temperature of the main cabin and the temperature of the motor are same dimension variables, and the change trends are similar, the temperature of the main cabin in the AUV motor and the temperature of the motor are taken as fusion variables in a leakage monitoring mode, and further, the key factors related to leakage, including depth, air pressure, cabin body deformation and main cabin temperature, are finally determined.
B, cabin deformation optimal estimation based on Kalman filtering:
because the deformation of the cabin body is obtained by a plurality of deformation sensors attached to the AUV cabin body, the deformation sensors can amplify the measurement noise when amplifying the weak deformation of the cabin body. Moreover, the measured value may have a large drift value due to the rebound phenomenon of the cabin body, so that it is necessary to process the deformation data.
In this embodiment, a kalman filter model is established, and deformation data values of the cabin actually measured by a plurality of deformation sensors are used as observed quantities, and a deformation value estimated by a numerical computation model (the model is in a form of y ═ Lx + B) using a depth value is used as a predicted value. By utilizing the observed quantity and the predicted value and combining with a Kalman filtering equation, the optimal estimated value of the deformation of the cabin body can be obtained, and the specific analysis is as follows:
firstly, AUV cabin deformation data is used as observed quantity, a deformation value estimated by a depth value according to a numerical calculation model is used as a predicted value, Kalman filtering is used for correcting the predicted value in real time, and iterative calculation is carried out according to an expression (2-1) -an expression (2-6).
Figure BDA0003570314930000062
Zk＝LDk+B+Vk (2-2)
Pk -＝Pk-1+Q (2-3)
Figure BDA0003570314930000063
Figure BDA0003570314930000064
Figure BDA0003570314930000065
Wherein S iskIs an observed value of deformation of the cabin body at the moment k, Wk-1And VkRepresenting the Gaussian noise in the observation and prediction processes, respectively, with their covariances Q and R, ZkThe predicted value of the deformation at the time k, D, is obtained from the modelkIs the depth measurement, L and B are numerical calculation model parameters; p isk-1Is that
Figure BDA0003570314930000066
Covariance matrix of (2), PkIs that
Figure BDA0003570314930000067
The covariance matrix of (2).
Figure BDA0003570314930000068
Is that
Figure BDA0003570314930000069
Covariance matrix of, KkIs the gain of the kalman gain (in),
Figure BDA0003570314930000071
is the optimal deformation quantity at the last moment,
Figure BDA0003570314930000072
is the optimal deformation quantity at the current moment,
Figure BDA0003570314930000073
is the predicted value of k time obtained according to the optimal value of the previous time, and the embodiment will be described
Figure BDA0003570314930000074
Is set to 0, will
Figure BDA0003570314930000075
Pk，Pk-1The initial values of the covariance matrix are set as identity matrices so that the output of the filter after a period of iteration is independent of the initial values.
And obtaining a filtering value through the processing, and performing cyclic iteration on the data according to a time sequence based on the current filtering value so as to improve the reliability of the deformation data and reduce the false alarm rate. And (3) calculating the predicted value of the deformation quantity according to theoretical model reference by using the depth value of the deformation data in the AUV cabin, taking the strain acquisition value as observed quantity, establishing respective Kalman models, and solving the optimal deformation quantity of different positions at the current moment by using formula 2. And filtering the deformation data acquired by the deformation sensors at different positions, and averaging the deformation data to obtain the final optimal deformation of the cabin body so as to measure the overall deformation degree of the AUV cabin body.
Step C, finally, establishing a self-encoder through related data to carry out leakage anomaly detection;
and after the leakage sensor gives an alarm, inputting data such as air pressure, main cabin temperature, depth, cabin optimal deformation, initial temperature, initial air pressure and the like acquired by the cabin sensor into a leakage self-encoder to perform leakage abnormity detection. Because the temperature and the air pressure have larger influence along with the environment, human factors (such as vacuumizing) also have visual change on the air pressure, the change trend of the air pressure also changes along with the change of the initial quantity, the experimental data is the change of the presentation data, and the interception depth changes more, namely the task execution stage with rich variable changes. Thus, the initial cabin temperature and the initial air pressure are introduced as characteristic data, wherein the cabin air pressure and temperature at the time when the AUV enters water are taken as the initial cabin temperature and the initial air pressure.
As shown in fig. 2, the initial temperature and the air pressure are taken as references, the leakage self-encoder is trained by six variables, the self-encoder network structure, the learning rate of input data and the iteration times are adjusted to obtain the self-encoder structure with a smaller loss function, and the AUV depth, the cabin body optimal deformation, the main cabin temperature and the air pressure are taken as characteristic data of leakage abnormity. And constructing a data set by using the historical data in the cabin of the AUV, wherein 70% of the data set is used as a training set, 15% of the data set is used as a verification set, and 15% of the data set is used as a test set.
The constructed structure of the leakage self-encoder encodes and outputs four dimensions by using a two hidden layer MLP structure, compresses input six feature data, extracts related features, and reconstructs input data by using a decoder with a symmetrical structure. As in fig. 3, the encoder input layer contains 6 input cells, the hidden layer 1 contains 64 neurons and the hidden layer 2 contains 16 neurons. The hidden layer 2 outputs 4 feature quantities. As a feature quantity compressed from 6 input quantities. While these 4 feature quantities are used as input quantities in the hidden layer 1 in the decoder. Meanwhile, the hidden layer 1 in the decoder comprises 16 neurons, and the hidden layer 2 in the decoder comprises 64 neurons. The final output layer contains 6 output features, i.e. the reconstruction volume. The learning rate of the whole self-encoder is 0.0001, the iteration times are 1000, and Adam is used as an optimization function. The activation function adopts RELU, which solves the problems of gradient explosion and gradient disappearance to a great extent.
The input of the leakage self-encoder is six variables, the learned characteristics of the data are characteristics between the data, and the reconstruction error of each data is caused by the abnormal change of each data, so the maximum value of the sum of the absolute values of the reconstruction errors of 6 variables in the test set is used as a judgment threshold, as shown in formula 3, e' is the reconstruction error of each variable under the leakage data, e represents the reconstruction errors of different variables under the normal state, and subscript i sequentially refers to six input data.
Figure BDA0003570314930000081
Therefore, the self-encoder takes the six variables (air pressure, main cabin temperature, depth, optimal cabin deformation value, initial temperature and initial air pressure) as input layers, and after data features are extracted through the middle two hidden layers, six reconstruction values are output on the output layers, wherein the six reconstruction values are respectively as follows: the method comprises the steps of reconstructing AUV depth, reconstructing optimal deformation of a cabin body, reconstructing main cabin temperature, reconstructing air pressure, reconstructing initial cabin temperature and reconstructing initial air pressure, subtracting reconstructed values from input values to obtain corresponding residual errors, summing absolute values of the residual errors to obtain judgment parameters, and judging that leakage is abnormal when the judgment parameters are larger than a set judgment threshold value, or judging that leakage is misjudged if the judgment parameters are not larger than the set judgment threshold value.
In addition, it should be noted that, in the embodiment, in the step B, when the deformation quantity of the cabin is estimated, the method is not limited to the kalman filtering described in the embodiment, and other filtering algorithms such as the extended kalman filtering, the unscented kalman filtering, the particle filtering, and other optimal estimation methods are also feasible; and step C may also utilize other types of self-encoders, such as convolutional self-encoder, cyclic neural network self-encoder, sparse self-encoder, stacked self-encoder, etc., which are not described herein too much.
The above description is only a preferred embodiment of the present invention, and not intended to limit the present invention in other forms, and any person skilled in the art may apply the above modifications or changes to the equivalent embodiments with equivalent changes, without departing from the technical spirit of the present invention, and any simple modification, equivalent change and change made to the above embodiments according to the technical spirit of the present invention still belong to the protection scope of the technical spirit of the present invention.

Claims (4)

1. An auxiliary method for leakage monitoring in an AUV cabin, comprising the steps of:
step A, analyzing the correlation between each variable and leakage based on AUV (autonomous underwater vehicle) cabin data to determine a key factor most correlated to cabin leakage;
the variables comprise air pressure, depth, main cabin temperature, main cabin humidity, motor temperature, cabin body deformation, system battery voltage, system battery current, power battery voltage and power battery current, and the determined key factors comprise the depth, the air pressure, the cabin body deformation and the main cabin temperature;
b, carrying out optimal estimation on deformation of the cabin based on Kalman filtering;
taking cabin deformation data values actually measured by a plurality of deformation sensors on the AUV cabin as observed values, taking deformation values estimated according to a numerical computation model by using depth values as predicted values, and estimating the optimal deformation of the cabin by combining a Kalman filtering equation based on the observed values and the predicted values;
step C, establishing a leakage self-encoder through related data to carry out leakage abnormity monitoring;
(1) constructing a data set by using historical cabin data of the AUV, wherein the data set comprises six variables of air pressure, main cabin temperature, depth, optimal cabin deformation, initial temperature and initial air pressure, and determining a training set, a verification set and a test set;
(2) constructing and training a leakage self-encoder, taking six variables of test concentration depth, main cabin temperature, air pressure, cabin deformation optimal value, initial cabin temperature and initial air pressure as input, and outputting corresponding 6 reconstruction values after being processed by the leakage self-encoder; the difference value between the reconstruction value and the input is a reconstruction error, and the maximum value of the sum of the absolute values of the reconstruction errors of 6 variables in the test set is used as a judgment threshold;
(3) when the AUV has a leakage sensor to report leakage alarm, inputting the current air pressure, the temperature and the depth of the main cabin, the optimal deformation of the cabin, the initial temperature and the initial air pressure data into a trained leakage self-encoder to obtain the sum of the absolute values of the reconstruction errors of 6 variables at the current moment, recording the sum as a judgment parameter, when the judgment parameter is greater than a judgment threshold value, considering that the leakage is abnormal, otherwise, judging that the leakage is misjudged.
2. Auxiliary method for leakage monitoring within an AUV capsule according to claim 1, characterized in that: in the step A, when the key factor is determined, the key parameter related to the air pressure is determined through correlation analysis:
(1) the correlation of the air pressure with other variables was calculated using spearman correlation coefficients:
Figure FDA0003570314920000011
where Ps is a correlation calculated from the spearman correlation coefficient, x represents the level of the air pressure, y represents the level of any one of the other nine variables, i represents the ith variable data, n represents the total number of the variable data,
Figure FDA0003570314920000012
represents the rank average of x and y;
(2) carrying out hypothesis test on the correlation to provide a primary hypothesis H0For Ps 0, assume H1Ps is not equal to 0; calculating statistic z obeying standard normal distribution by taking original hypothesis establishment as background*；
Figure FDA0003570314920000021
Calculating cumulative distribution function of standard normal distribution, and calculating cumulative distribution function valuez*And (3) substituting a certain corresponding determined point m into the following formula to obtain a check value p:
p＝2(1-m)
if p is>0.05, the original hypothesis cannot be rejected at a confidence level of 95%, i.e. the original hypothesis H is satisfied0No significant difference exists, and no correlation exists among data; otherwise, the original hypothesis is rejected at a 95% confidence level, i.e., hypothesis H is satisfied1And the data have correlation.
3. Auxiliary method for leakage monitoring within an AUV capsule according to claim 1, characterized in that: the step B is specifically realized by the following steps:
(1) firstly, AUV cabin deformation data is used as observed quantity, a deformation value estimated by using a depth value according to a numerical calculation model y, namely Lx + B, is used as a predicted value, Kalman filtering is used for correcting the predicted value in real time, and iterative calculation is as follows:
Figure FDA0003570314920000022
Zk＝LDk+B+Vk
Figure FDA0003570314920000023
Figure FDA0003570314920000024
Figure FDA0003570314920000025
Figure FDA0003570314920000026
wherein S iskIs an observed value of deformation of the cabin body at the moment k, Wk-1And VkRepresenting the Gaussian noise in the observation and prediction processes, respectively, and the corresponding covariances are Q and R, ZkThe predicted value of the deformation quantity at the k moment obtained according to the model, DkIs a depth measurement, L and B are numerical calculation model parameters; p isk-1Is that
Figure FDA0003570314920000027
Covariance matrix of (2), PkIs that
Figure FDA0003570314920000028
The covariance matrix of (a) is determined,
Figure FDA0003570314920000029
is that
Figure FDA00035703149200000210
Covariance matrix of, KkIs the gain of the kalman gain (in),
Figure FDA00035703149200000211
is the optimal deformation quantity at the last moment,
Figure FDA00035703149200000212
is the optimal deformation quantity at the current moment,
Figure FDA00035703149200000213
the predicted value of the k moment is obtained according to the optimal value of the previous moment; wherein
Figure FDA00035703149200000214
Is set to 0, will
Figure FDA00035703149200000215
Figure FDA00035703149200000216
Pk，Pk-1Setting the initial value of the equal covariance matrix as a unit matrix;
(2) obtaining a filtering value through the processing of the step (1), carrying out cyclic iteration on the data according to a time sequence based on the current filtering value, and solving the optimal deformation quantity of the deformation sensors at different positions at the current moment;
(3) after the deformation data collected by the deformation sensors at different positions are filtered, the average deformation data is taken as the final optimal deformation of the cabin body, and the overall deformation degree of the AUV cabin body is measured.
4. Auxiliary method for leakage monitoring within an AUV capsule according to claim 1, characterized in that: in the step C, the leakage self-encoder includes an input layer, an encoder, a decoder, and an output layer, the encoder and the decoder are structurally symmetric and respectively include two hidden layers, the two hidden layer MLP structures of the encoder encode and compress six input feature data, extract relevant features of four dimensions of output depth, cabin optimal deformation, main cabin temperature, and air pressure, and meanwhile, the 4 feature quantities are used as input quantities of the decoder, and finally, the output layer outputs 6 output features, that is, reconstruction quantities.
CN202210320437.8A 2022-03-29 2022-03-29 Auxiliary method for leakage monitoring in AUV (autonomous Underwater vehicle) cabin Pending CN114662327A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210320437.8A CN114662327A (en) 2022-03-29 2022-03-29 Auxiliary method for leakage monitoring in AUV (autonomous Underwater vehicle) cabin

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210320437.8A CN114662327A (en) 2022-03-29 2022-03-29 Auxiliary method for leakage monitoring in AUV (autonomous Underwater vehicle) cabin

Publications (1)

Publication Number Publication Date
CN114662327A true CN114662327A (en) 2022-06-24

Family

ID=82032923

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210320437.8A Pending CN114662327A (en) 2022-03-29 2022-03-29 Auxiliary method for leakage monitoring in AUV (autonomous Underwater vehicle) cabin

Country Status (1)

Country Link
CN (1) CN114662327A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116930626A (en) * 2023-09-18 2023-10-24 江苏省电子信息产品质量监督检验研究院(江苏省信息安全测评中心) Low-frequency electromagnetic field exposure monitoring system and method
CN117094563A (en) * 2023-10-16 2023-11-21 南京六季光电技术研究院有限公司 Intelligent liquid waste leakage monitoring system and method based on big data

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116930626A (en) * 2023-09-18 2023-10-24 江苏省电子信息产品质量监督检验研究院(江苏省信息安全测评中心) Low-frequency electromagnetic field exposure monitoring system and method
CN116930626B (en) * 2023-09-18 2023-12-05 江苏省电子信息产品质量监督检验研究院(江苏省信息安全测评中心) Low-frequency electromagnetic field exposure monitoring system and method
CN117094563A (en) * 2023-10-16 2023-11-21 南京六季光电技术研究院有限公司 Intelligent liquid waste leakage monitoring system and method based on big data
CN117094563B (en) * 2023-10-16 2024-02-06 南京六季光电技术研究院有限公司 Intelligent liquid waste leakage monitoring system and method based on big data

Similar Documents

Publication Publication Date Title
CN114757309B (en) Multi-physical-field monitoring data collaborative fusion engineering disaster early warning method and system
CN114662327A (en) Auxiliary method for leakage monitoring in AUV (autonomous Underwater vehicle) cabin
EP1416348B1 (en) Methodology for temporal fault event isolation and identification
Cheung et al. The application of statistical pattern recognition methods for damage detection to field data
CN110555247A (en) structure damage early warning method based on multipoint sensor data and BilSTM
CN112785091A (en) Method for performing fault prediction and health management on oil field electric submersible pump
CN114936520B (en) Steam turbine fault early warning analysis method based on improved MSET
Moeinifard et al. Lost vibration test data recovery using convolutional neural network: a case study
CN114091525A (en) Rolling bearing degradation trend prediction method
CN113011248A (en) Nuclear power plant fault diagnosis method based on LSTM
CN111753776A (en) Structural damage identification method based on echo state and multi-scale convolution combined model
CN117809164A (en) Substation equipment fault detection method and system based on multi-mode fusion
CN115270266A (en) Slope monitoring state risk judgment method based on improved analytic hierarchy process
CN116662920B (en) Abnormal data identification method, system, equipment and medium for drilling and blasting method construction equipment
CN114021620A (en) Electrical submersible pump fault diagnosis method based on BP neural network feature extraction
CN113742983A (en) Long-span structural damage identification method based on depth self-encoder neural network
CN117470529A (en) Static friction fault detection method and system for valve of process control system
CN115750341A (en) Roots compressor fault diagnosis method and system
CN115358294A (en) Micro fault detection method for high-speed train traction system
CN115311238A (en) Photovoltaic module defect damage identification method based on image analysis
CN112069724A (en) Rocket health degree evaluation method based on long-time and short-time memory self-encoder
CN115235877A (en) Corrosion state monitoring system and stress corrosion state evaluation method and device
CN108133287B (en) Ionosphere anomaly detection method based on Kalman filtering typhoon process
CN112098066A (en) High-voltage shunt reactor fault diagnosis method and system based on gate control circulation unit
CN118090078B (en) Leakage online monitoring method for closed circulation water cooling system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination