Underwater radiated noise online prediction method based on hull vibration and self-noise

Abstract

The application discloses an online prediction method for underwater radiation noise based on body vibration and self-noise, and belongs to the technical field of ship noise monitoring and evaluation. The method is characterized in that: a plurality of test working conditions under different working conditions are designed according to a corresponding navigation body, and parameters such as body vibration, self-noise and underwater radiation noise of the navigation body under the test working conditions are accurately obtained; based on the above-mentioned collected parameters, a transfer function and a working condition clustering center under each working condition are accurately calculated; parameter information of the navigation body in an actual operation process is collected in real time, and underwater radiation noise is accurately predicted according to the transfer function and the working condition clustering center. The application provides an online prediction method for underwater radiation noise based on body vibration and self-noise, can reduce nonlinearity and ill-conditioned property of solving of underwater radiation noise transfer function under influence of external disturbance, accurately calculate the transfer function, and enable underwater radiation noise to be accurately obtained in real time under different running environments and running states of the navigation body.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of ship noise monitoring and evaluation, and particularly relates to an online prediction method for underwater radiation noise based on body vibration and self-noise. BACKGROUND

[0002] Ship radiation noise is mainly composed of mechanical noise, propeller noise and hydrodynamic noise, and is a complex multi-source coupled noise source. Real-time prediction of ship self-radiation noise level to guide the reasonable implementation of ship vibration and noise reduction measures is of great significance to improve the comprehensive acoustic performance of the ship.

[0003] In the prior art, the transfer path analysis method is mainly adopted to establish a "response-transfer path-response" model, the measured ship self-noise or vibration acceleration response is taken as the input, the far-field radiation noise response is taken as the output, and the transfer function between the input and output responses is solved in advance to predict the underwater radiation noise of the submarine in real time. However, due to the different excitation points, characteristic frequencies and amplitude sizes of the ship mechanical equipment, propeller and flow-induced noise under different operating conditions, the stability and robustness of the transfer function are poor, and the system is seriously ill, resulting in poor stability and robustness of the transfer function. And subject to the radiation noise test conditions, the number of test conditions is small, resulting in large dispersion between the operating conditions, and the response input often cannot meet the linear system assumption conditions required by the transfer path analysis method. At this time, the system transfer matrix calculated based on all the test data responses of the operating conditions cannot be applied to different types of operating conditions, resulting in large calculation error of the radiation noise in some frequency bands.

[0004] In view of this problem, domestic and foreign scholars have carried out a lot of research from different angles, and have introduced principal component analysis method, singular value decomposition method, regularization method and other methods from the experience of geophysical field, for optimizing the ill-conditioned problem caused by the transfer function solving, but there is no effective solution to the nonlinear problem in the solving process. SUMMARY

[0005] In view of one or more of the above defects or improvement needs of the prior art, the application provides an online prediction method for underwater radiation noise based on body vibration and self-noise, which can reduce the nonlinearity and ill-conditioning of the underwater radiation noise transfer function solving under the influence of external disturbance, to accurately calculate the transfer function, and further enable the underwater radiation noise to be accurately obtained in real time when the vehicle is in different running environments and running states.

[0006] In order to achieve the above object, the application provides an online underwater radiation noise prediction method based on body vibration and self-noise, which is used for predicting the underwater radiation noise of a vehicle in real time and accurately according to the body vibration and self-noise of the vehicle, and comprises the following steps:

[0007] S1, according to the actual navigation requirement of the vehicle, designing several underwater radiation noise test working conditions of the vehicle, and collecting the body vibration and self-noise X, the underwater radiation noise Y and the navigation parameters of the vehicle in each test working condition;

[0008] S2, constructing an original working condition clustering feature W according to the body vibration and self-noise X and the navigation parameters;

[0009] Wherein, the original working condition clustering feature W comprises a frequency domain feature set w fre , the frequency domain feature set w fre comprises a spectrum centroid E SC and a spectrum centroid extension feature E SS ;

[0010] S3, determining a sensor confidence P α of a self-noise sensor in each working condition, weighting the original working condition clustering feature W by using the sensor confidence P α to obtain a weighted feature matrix W weigh , and extracting a feature matrix W opt of a main working condition cluster through the weighted feature matrix W weigh ;

[0011] S4, performing working condition clustering on the feature matrix W opt to form K working conditions and a clustering center M;

[0012] S5, calculating a transfer function T i of the body vibration and self-noise X i to the underwater radiation noise Y i in each working condition of the K working conditions, i.e. i=1, 2, 3, …, K, and saving a transfer function library T;

[0013] S6, measuring the body vibration and self-noise X test and the navigation parameters v test of the vehicle during navigation in real time, and calculating a working condition clustering feature W test during navigation;

[0014] S7, performing working condition matching on the working condition clustering feature W test during navigation and the clustering center M, and selecting a radiation noise transfer function T test during the navigation from the transfer function library T according to the matching result.

[0015] S8, the body vibration and self-noise X during the voyage test and the radiated noise transfer function T during the voyage test , calculate the actual underwater radiated noise Y of the vehicle during the voyage test .

[0016] As a further preferred embodiment of the present application, the original working condition clustering feature W further comprises a set of voyage features w sai , a set of time domain features w time , a set of statistical domain features w sta .

[0017] As a further preferred embodiment of the present application, the sensor confidence P α is calculated by the ratio of the number of working condition samples accurately identified by the working condition clustering feature of the vehicle under a single sensor to the number of working condition samples.

[0018] As a further preferred embodiment of the present application, the sensor confidence P α is calculated by the following formula:

[0019]

[0020] In the formula, the sensor confidence at different positions, N con is the number of working condition samples, n i is the number of regions divided by the sensor, is the number of working condition samples accurately identified based on the feature information of a single sensor.

[0021] As a further preferred embodiment of the present application, the S3 comprises the following steps:

[0022] S31, the weighted feature matrix W weigh is decentered, and the average value of the working condition sample features is calculated

[0023] S32, the respectively from the covariance matrix G X and G Y in X direction and Y direction two dimensions are eigenvalue decomposition, get feature matrix U X and U Y , and the eigenvalue δ i , it satisfies δ i+1 ≥ δ i ; and assuming that the first L principal component components are selected;

[0024] S33, the kth working condition sample feature matrix after dimension reduction is selected as the feature matrix

[0025] As a further preferred embodiment of the present application, in the S31, the average value of the working condition sample feature is calculated according to the following formula:

[0026]

[0027] wherein k is the number of working condition samples (k = 1, 2, …, N g ), is the weighted feature matrix under the kth working condition.

[0028] As a further preferred embodiment of the present application, in the S33, the main feature matrix W opt may be expressed as:

[0029]

[0030] wherein, U Yr and U Xr are the first L eigenvectors of the feature matrix U Y and U X , respectively.

[0031] As a further preferred embodiment of the present application, the S4 comprises the following steps:

[0032] S41, based on the neighborhood sample tightness, the initial clustering center selection is optimized to form a set of K initial clustering centers m = {m1, m2, …, m K};

[0033] S42, the feature matrix W opt is weighted based on the sample working condition feature dispersion, and the Euclidean distance d(x i , x j ) between the weighted working condition sample features is used as the clustering measurement criterion for representing the similarity between samples, and the Euclidean distance matrix d between all samples is calculated.

[0034] S43, the intra-class-inter-class distance clustering evaluation criterion E' is used as the target function of the clustering implementation, and the clustering implementation is terminated when the clustering implementation result meets the target function, and K classes of working condition samples and clustering centers M are formed.

[0035] As a further preferred embodiment of the present application, the S5 comprises the following steps:

[0036] S51, assuming that the number of the ith class of working condition samples is r i ;

[0037] S52, the transfer function T i is calculated by using a hybrid regularization method.

[0038] S53, store the transfer function T in the transfer function library T i to the transfer function library T, and repeat steps S51 and S52 until the transfer function library T is completely stored with the transfer functions T under all working conditions i , and preset in the system.

[0039] As a further preferred embodiment of the present application, the hybrid regularization method calculates the transfer function T i The calculation formula includes:

[0040]

[0041] wherein T reg is the regularization initial value of the solution, indicates the hybrid regularization parameter, which can be solved by BL-curve, that is, a plurality of points are intercepted near the maximum curvature point of the L curve, the horizontal and vertical coordinates can be represented as (x, y), and the optimal hybrid regularization parameter is calculated by B-spline interpolation The calculation formula is as follows:

[0042]

[0043] wherein x' and y' represent the first derivative coordinates of the curve after B-spline interpolation, and x" and y" represent the second derivative coordinates of the curve after interpolation.

[0044] The above improved technical features can be combined with each other as long as they do not conflict with each other.

[0045] Overall, compared with the prior art, the above technical solutions conceived by the present application have the following beneficial effects:

[0046] (1) The online prediction method for underwater radiated noise based on body vibration and self-noise in the present application can accurately obtain the body vibration and self-noise, underwater radiated noise and other parameters of the vehicle under the test working condition by designing a plurality of different test working conditions for the vehicle, and can accurately calculate the transfer function and working condition clustering center under similar working conditions based on the above collected parameters, wherein the radiated noise transfer function under each working condition is approximately linear. Moreover, by collecting the sensor confidence of the vehicle body vibration and self-noise sensor, the collection and application of the navigation feature set, the spectrum centroid and the spectrum centroid extension feature, the accuracy and anti-interference ability of the feature matrix are effectively improved, thereby ensuring the accuracy of the working condition clustering, reducing the nonlinearity of the underwater radiated noise transfer function solving, and enabling the vehicle to accurately obtain the underwater radiated noise in real time under different operating environments and operating states.

[0047] (2) The online prediction method for underwater radiation noise based on body vibration and self-noise has the advantages that: the initial clustering center selection is optimized based on the neighborhood sample tightness, so that the initial clustering center can be accurately determined when the working condition clustering is performed, and the operation efficiency of the working condition clustering is effectively improved; the feature matrix is weighted based on the sample working condition feature dispersion, so that the features in the feature matrix can be quickly and accurately clustered when the working condition clustering is performed, and the operation efficiency and accuracy of the working condition clustering are further improved; and when the transfer function is calculated by using the hybrid regularization method, a plurality of points near the maximum curvature point of the L curve are intercepted, and the optimal hybrid regularization parameter is calculated by using the B-spline interpolation, so that the nonlinearity in the transfer function solving process is avoided, and the solving accuracy and efficiency of the transfer function are improved as much as possible, and the working efficiency of the online prediction method in actual use is greatly improved.

[0048] (3) The online prediction method for underwater radiation noise based on body vibration and self-noise has the advantages of high prediction accuracy, high working efficiency and simple process, which accurately collects the parameters such as body vibration and self-noise, underwater radiation noise and the like in the test working condition, accurately calculates the transfer function between the body vibration and self-noise and the underwater radiation noise of the vehicle under different working conditions according to the parameters collected in the test working condition, and then accurately predicts the underwater radiation noise generated by the vehicle during operation by collecting the body vibration and self-noise of the vehicle. Moreover, the online prediction method designs the confidence of each sensor for collecting the body vibration and self-noise of the vehicle, so that the original working condition clustering feature can be more consistent with the body characteristics of the vehicle when the original working condition clustering feature is calculated, and the accuracy of the online prediction of the underwater radiation noise is ensured; meanwhile, the navigation feature set and the frequency domain feature set are added in the original working condition clustering feature, and the spectrum centroid and the spectrum centroid extension feature are added in the frequency domain feature set, so that the original working condition clustering feature can be as accurate as possible, and the accuracy of the online prediction of the underwater radiation noise of the vehicle is further improved, which has excellent economic benefits and promotion value. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 is the flow chart of the online prediction method for underwater radiation noise based on body vibration and self-noise in the application;

[0050] Figure 2 is the working condition sample clustering result chart of the online prediction method for underwater radiation noise based on body vibration and self-noise in the application under a certain working condition;

[0051] Figure 3 is the result chart of the radiation noise prediction radiation with frequency of the online prediction method for underwater radiation noise based on body vibration and self-noise in the application under a certain working condition;

[0052] Figure 4 is a result chart of absolute value of error of the underwater radiation noise prediction method based on body vibration and self-noise in the present application in a certain working condition. DETAILED DESCRIPTION

[0053] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0054] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only used to facilitate the description of the present application and simplify the description, and therefore cannot be understood as indicating or implying that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0055] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise explicitly specified and limited.

[0056] In the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection", "fixing" and the like should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integral; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, it can be the internal communication of two elements or the interaction relationship between two elements, unless otherwise explicitly limited. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0057] In the present application, unless otherwise explicitly specified and limited, the first feature is "on" or "under" the second feature can be that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, the first feature "above", "over" and "on" the second feature can be that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is horizontally higher than the second feature. The first feature "below", "under" and "under" the second feature can be that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is horizontally lower than the second feature.

[0058] Embodiments:

[0059] As shown in Figures 1-4 The online prediction method of underwater radiated noise based on body vibration and self-noise in the preferred embodiment of the present application can ensure that the use of transfer function to solve underwater radiated noise is always linear under the influence of external disturbance, and thus the underwater radiated noise can be accurately obtained in real time when the vehicle is in different operating environments and operating states.

[0060] Specifically, as shown in Figure 1 The online prediction method of underwater radiated noise based on body vibration and self-noise is used to accurately predict the underwater radiated noise of the vehicle in real time according to the body vibration self-noise of the vehicle, and includes the following steps:

[0061] S1, according to the actual navigation requirements of the vehicle, design several vehicle underwater radiated noise test working conditions, and collect the body vibration and self-noise X, underwater radiated noise Y, navigation parameters and equipment operating state of the vehicle in each test working condition;

[0062] Preferably, the navigation parameters include but are not limited to speed v, propeller speed R, diving depth d

[0063] S2, constructing original working condition clustering features W according to the body vibration and self-noise X and the navigation parameters;

[0064] Wherein, the original working condition clustering features W include but are not limited to the navigation feature set w sai , the time domain feature set w time , the frequency domain feature set w fre , the statistical domain feature set w sta , the frequency domain feature set w fre includes the spectrum centroid E SC and the spectrum centroid extension feature E SS ;

[0065] S3, determining the sensor confidence P α of the m self-noise sensors in each working condition αThe original working condition clustering feature W is weighted to obtain a weighted feature matrix W weigh , and the weighted feature matrix W weigh is obtained opt ;

[0066] S4, the feature matrix W opt is clustered according to working conditions to form K working conditions and clustering centers M;

[0067] S5, the transfer function T i of the body vibration and self-noise X i to the underwater radiation noise Y i is calculated under each of the K working conditions, i = 1, 2, 3, …, K, and a transfer function library T is saved;

[0068] S6, the body vibration and self-noise X test and the navigation parameters v test of the vehicle during navigation are measured in real time, and the working condition clustering feature W test during navigation is calculated;

[0069] S7, the working condition clustering feature W test during navigation is matched with the clustering center M, and the corresponding radiation noise transfer function T test (test ∈ (1, 2, 3, …, K)) is selected from the transfer function library T according to the matching result;

[0070] S8, the actual underwater radiation noise Y test of the vehicle during navigation is calculated according to the body vibration and self-noise X test during navigation and the radiation noise transfer function T test during navigation.

[0071] Preferably, the formula for calculating the actual underwater radiation noise Y test is:

[0072] Y test = X test T test

[0073] Further, in the preferred embodiment of the present application, in step S2, the navigation feature set w sai includes but is not limited to the speed index factor v e , the propeller speed index factor R e , and the diving depth index factor d e , which are calculated by using navigation parameters, and are specifically calculated by the following formula:

[0074] The speed index factor v e :

[0075] propeller rotation speed index factor R e :

[0076] submersion depth index factor d e :

[0077] wherein the parameters v max , R max , d max represent the maximum cruising speed of the vehicle, the maximum propeller rotation speed corresponding to the maximum cruising speed, and the maximum submersion depth, respectively.

[0078] The set of navigation characteristics w sai may be represented as follows:

[0079] w sai = {v e , R e , d e}

[0080] Further, the set of time domain characteristics w time includes, but is not limited to, the mean value w mean , the peak value w max , the variance w var , the peak-to-peak value w m-m , the mean amplitude w a-amp , the root amplitude w r-amp , the root mean square amplitude w m-amp , all of which are calculated from the body vibration and self-noise data X, specifically by the following formulas:

[0081] Mean value w mean :

[0082] Peak value w max : w max = max(|x(t)|)

[0083] Variance wvar:

[0084] Peak-to-peak value w m-m m-m: w = max(x(t))-min(x(t))

[0085] Mean amplitude w a-amp :

[0086] Root amplitude w r-amp :

[0087] Root mean square amplitude w m-amp :

[0088] wherein x(t) represents the collected time-domain signal (x(t) e X), and T represents the sampling duration of the signal x(t).

[0089] The time-domain feature set w time may be represented as follows:

[0090] w time = {w mean , w max , w var , w m-m , w a-amp , w r-amp , w m-amp}

[0091] Further, the frequency-domain feature set w fre also includes the one-sixth-octave energy E(f i ), the high-to-low frequency energy ratio E c (f), and the total energy E f , all of which are calculated from the body vibration and self-noise data X, and specifically include the following calculation formulas:

[0092] The one-sixth-octave energy E(fi) is:

[0093] The total energy E f is:

[0094] The high-to-low frequency energy ratio E c (f) is:

[0095] The spectral centroid E SC is:

[0096] The spectral centroid spread feature E SS is:

[0097] wherein E(f) represents the signal spectral level, f is the signal frequency, f h,i and f l,i represent the upper and lower limit frequencies of the i-th one-sixth-octave, respectively, E h (f) is the signal high-frequency energy, and E l (f) is the signal low-frequency energy.

[0098] The frequency-domain feature set w fre may be represented as follows:

[0099] w fre = {E(f i ), E c(f), E f ,E SC ,E SS}

[0100] Further, the statistical domain feature set w sta including but not limited to a peak factor w pea , a waveform factor w wav , a pulse factor w pul , a kurtosis index w kur and a skewness index w inc , are all calculated from the body vibration and self-noise data X, with the following formulas included:

[0101] The peak factor w pea : w pea = w max / w m-amp

[0102] The waveform factor w wav : w wav = w m-amp / w a-amp

[0103] The pulse factor w pul : w pul = w max / w a-amp

[0104] The kurtosis index w kur :

[0105] The skewness index w inc :

[0106] The statistical domain feature set w sta may be represented as:

[0107] w sta = {w pea , w wav , w pul , w kur , w inc}

[0108] Further preferably, the original working condition clustering feature W can be represented as:

[0109] W = [w sai , w time , w fre , w sta ]

[0110] Further, in the preferred embodiment of the present application, in step S3, the sensor confidence P αThe original working condition clustering feature W is weighted, which can help to accurately extract the key features in the original working condition clustering feature W, thereby effectively reducing the calculation amount in the working condition clustering process, improving the working efficiency, and greatly improving the accuracy of working condition clustering. The sensor confidence P α is accurately divided by analyzing the position of the sensor in the spacecraft and the comprehensive analysis of the equipment condition around the sensor position.

[0111] Specifically, the sensor confidence P α The original working condition clustering feature W is weighted, and the weighted feature W weigh can be represented as:

[0112] W weigh = P α W

[0113] Further, the sensor confidence P α is a weight coefficient matrix, which is calculated by the ratio of the number of working condition samples accurately identified by the spacecraft under a single sensor to the number of working condition samples. The calculation formula is as follows:

[0114]

[0115] In the formula, The sensor confidence at different positions is N con , the number of working condition samples is n i , the number of regions divided by the sensor is n , and the number of working condition samples accurately identified based on the feature information of a single sensor is N

[0116] More specifically, in the preferred embodiment of the present application, the feature matrix W opt of the main working condition clustering is extracted from the weighted original working condition clustering feature W weigh including the following steps:

[0117] S31, the weighted feature matrix W weigh is decentered, and the average value of the working condition sample features is calculated

[0118] Preferably, the average value of the working condition sample features is calculated including the following formula:

[0119]

[0120] where k is the number of working condition samples (k = 1, 2, …, N g ), is the weighted feature matrix under the kth working condition.

[0121] S32, respectively, and G X and G Y are eigenvalue decomposed to obtain eigenmatrix U X and U Y , and eigenvalue δ i satisfying δ i+1 ≥ δ i . Assuming that the first L principal components are selected, the cumulative contribution rate CCR(L) can be expressed as:

[0122]

[0123] where N w is the number of features of each sensor.

[0124] S33, the k-th feature matrix of the reduced dimensionality is selected as the feature matrix

[0125] Preferably, the main feature matrix W opt extracted can be expressed as:

[0126]

[0127] wherein, U Yr and U Xr are the first L eigenvectors of the eigenmatrix U Y and U X .

[0128] Further, in the preferred embodiment of the present application, in step S4, the feature matrix W opt is clustered by the D 2 -Kmeans clustering method to form K working conditions and clustering centers M, specifically including the following steps:

[0129] S41, based on the neighborhood sample tightness, the initial clustering center selection is optimized to form a set of K initial clustering centers m={m1, m2, …, m K};

[0130] S42, the feature matrix W opt is weighted based on the sample working condition feature dispersion, and the Euclidean distance d(x i , x j ) between the weighted working condition sample features is used as the clustering measurement criterion to represent the similarity between samples, and the Euclidean distance matrix d between all samples is calculated.

[0131] S43, taking the intra-class-inter-class distance clustering evaluation criterion E' as a target function of clustering implementation, terminating when the clustering implementation result meets the target function, and forming K classes of working condition samples and clustering centers M.

[0132] In step S41, the initial clustering center selection based on neighborhood sample tightness optimization is adopted, so that the initial clustering center can be accurately determined when the working condition clustering is performed, and the operation efficiency of the working condition clustering is effectively improved. Moreover, in step S42, the feature matrix W opt is weighted, so that the features in the feature matrix W opt can be quickly and accurately clustered when the working condition clustering is performed, and the operation efficiency of the working condition clustering is further improved.

[0133] Further, in the preferred embodiment of the present application, step S5 comprises the following steps:

[0134] S51, assuming that the number of the i-th class of working condition samples is r i , the transfer function T i can be represented as:

[0135]

[0136] wherein, represents a set of m body vibration self-noise sensor data under r working conditions of the i-th class, is a generalized inverse matrix of , and is a set of radiation noise data under r working conditions of the i-th class.

[0137] S52, the transfer function T i is calculated by using a hybrid regularization method.

[0138] Since the number of working conditions r is much smaller than the number of sensor measuring points m in the process of calculating the transfer function T i , the equation set is not suitable, and a pathological problem will occur in the solving process. In order to avoid the pathological problem in the process of calculating the transfer function T i , in the preferred embodiment of the present application, the hybrid regularization method is used to calculate the transfer function T i , and specifically, the hybrid regularization calculation formula is as follows:

[0139]

[0140] wherein, T reg is a regular initial value of the solution, and The mixed regular parameter is represented, and the BL-curve method can be solved by interpolation optimization, that is, a plurality of points are intercepted near the maximum curvature point of the L curve, the horizontal and vertical coordinates can be represented as (x, y), and the optimal mixed regular parameter is calculated by B-spline interpolation The calculation formula is as follows:

[0141]

[0142] Wherein, x' and y' represent the first derivative coordinates of the curve after B-spline interpolation, and x" and y" represent the second derivative coordinates of the curve after interpolation.

[0143] S53, the transfer function T i is saved into the transfer function library T, and steps S51 and S52 are repeated until the transfer function library T completely stores the transfer functions T under all working conditions i , and it is preset in the system.

[0144] Further preferably, in the preferred embodiment of the present application, the prediction method can be divided into two stages in actual use for a certain vehicle, the first stage is mainly to realize the calculation of the special transfer function and the working condition clustering of the vehicle by adopting steps S1-S5, and the second stage is to realize the accurate prediction of the underwater radiation noise of the vehicle during the navigation period by adopting steps S6-S8.

[0145] The online prediction method of underwater radiation noise based on body vibration and self-noise in the application has high prediction accuracy, high work efficiency and simple process, which accurately collects the parameters of body vibration, self-noise and underwater radiation noise in the test working condition, and accurately calculates the transfer function between the body vibration and self-noise and the underwater radiation noise of the vehicle under different working conditions through the parameters collected in the test working condition, so that the underwater radiation noise generated by the vehicle during the navigation can be accurately predicted by collecting the body vibration and self-noise of the vehicle. Moreover, through the design of the confidence of the sensor for collecting the body vibration and self-noise in the vehicle, it can be ensured that the original working condition clustering characteristics can be more consistent with the body characteristics of the vehicle, thereby ensuring the accuracy of the online prediction of the underwater radiation noise. At the same time, by adding the navigation feature set and the frequency domain feature set in the original working condition clustering characteristics, and adding the spectrum centroid and the spectrum centroid extension feature in the frequency domain feature set, it is ensured that the original working condition clustering characteristics can be as accurate as possible, and the accuracy of the online prediction of the underwater radiation noise of the vehicle is further improved, which has excellent economic benefit and popularization value.

[0146] Those skilled in the art can easily understand that the above description is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An online prediction method of underwater radiated noise based on body vibration and self-noise, which is used to accurately predict the underwater radiated noise of a vehicle in real time according to the body vibration and self-noise of the vehicle, characterized in that, comprising the steps of: S1, according to the actual navigation requirements of the vehicle, design several underwater radiation noise test working conditions of the vehicle, and collect the body vibration and self-noise in each test working condition , underwater radiation noise , navigation parameters of the vehicle; S2, constructing the original working condition clustering feature according to the body vibration and self-noise and the navigation parameter ; The original working condition clustering features The frequency domain feature set The frequency domain feature set The spectrum centroid The spectrum centroid extension feature ; S3, determining sensor confidence of the vibration and self-noise sensors in each working condition , using the sensor confidence to the original working condition clustering features weighting, obtaining a weighted feature matrix , and extracting a feature matrix of the main working condition cluster through the weighted feature matrix ;​ S4, to the feature matrix S5, performing working condition clustering to form Working condition classes and cluster centers ; S5, calculating the vibration and self-noise of the body in each of the classes of working conditions to underwater radiated noise transfer function , and saving the obtained transfer function library ; S6, measuring body vibration and self-noise of the vehicle during navigation in real time with navigation parameters and calculating the working condition clustering features during navigation ; S7, clustering the working condition features during the voyage with the cluster centers performing working condition matching, and selecting the radiation noise transfer function during the voyage from the transfer function library according to the result of the working condition matching ;​ S8, calculating actual underwater radiated noise of the vehicle during the voyage based on the body vibration and self-noise during the voyage and the radiated noise transfer function during the voyage .​ 2. The online prediction method of underwater radiated noise based on body vibration and self-noise according to claim 1, wherein, The original working condition clustering feature Also includes a navigation feature set , time domain feature set , statistical domain feature set .

3. The online prediction method of underwater radiated noise based on the body vibration and self-noise according to claim 1, wherein, The sensor confidence The number of working condition samples is accurately identified by clustering features of working conditions under a single sensor by the aircraft.

4. The online prediction method of underwater radiated noise based on the body vibration and self-noise according to claim 3, wherein, The sensor confidence The formula for calculating the sensor confidence is as follows: In the formula, Sensor confidence at different locations, is the number of working condition samples, is the number of regions delineated by the sensor, is the number of working condition samples accurately identified based on single sensor feature information.

5. The online prediction method of underwater radiated noise based on body vibration and self-noise according to any one of claims 1-4, wherein, said S3 comprises the steps of: S31、to the weighted feature matrix Decentralization, computing the average value of the working condition sample features ; S32、to respectively from the covariance matrix of two dimensions of X direction and Y direction and Eigenvalue decomposition is carried out, and the characteristic matrix and , and the eigenvalue , which satisfies ; and assuming that the first principal component components are selected; S33, select the reduced dimensionality of the first condition sample feature matrix as the feature matrix .

6. The online prediction method of underwater radiated noise based on the body vibration and self-noise according to claim 5, wherein, In the S31, the average value of the working condition sample features including the following equation: wherein, is the number of working condition samples , is the weighted feature matrix under the th working condition.

7. The online prediction method of underwater radiated noise based on the body vibration and self-noise according to claim 5, wherein, In the S33, the feature matrix may be expressed as: in, , and The characteristic matrix is ​​respectively and The former 1 eigenvector.

8. The online prediction method of underwater radiated noise based on body vibration and self-noise according to any one of claims 1-4, 6, 7, wherein, said S4 comprises the steps of: S41, based on the neighborhood sample tightness optimization initial clustering center selection, form a set of initial clustering centers ; S42、based on the sample working condition characteristic discrete degree to the characteristic matrix weighted, and the weighted working condition sample characteristics between the Euclidean distance For the measurement criterion of clustering, used to represent the similarity between samples, calculate the Euclidean distance matrix between all samples ; S43, intra-inter class distance clustering evaluation criterion as a target function for the clustering implementation, the clustering implementation is terminated when the clustering implementation result conforms to the target function, forming class condition samples and clustering centers .

9. The online prediction method of underwater radiated noise based on the body vibration and self-noise according to any one of claims 1-4, 6, 7, wherein, said S5 comprises the steps of: S51, assuming the first The number of class condition samples is ; S52, calculate the transfer function using the hybrid regularization method ; S53, save said transfer function to a transfer function library and repeat steps S51, S52 until said transfer function library is completely stored with transfer functions for all operating conditions after which it is pre-installed in the system.

10. The online prediction method of underwater radiated noise based on body vibration and self-noise according to claim 9, wherein, The hybrid regularization method computes the transfer function of the computational formula comprising: wherein, is the regular initial value of the solution, represents the mixed regular parameter, which can be solved by the BL-curve, that is, a plurality of points are intercepted near the maximum curvature point of the L curve, and the horizontal and vertical coordinates can be represented as , and the optimal mixed regular parameter is calculated by B-spline interpolation , and the calculation formula is as follows: wherein, and denotes the first derivative coordinate of the interpolated curve, and denotes the second derivative coordinate of the interpolated curve.

Images