A Fast Identification and Screening Method for the Characteristic Line Spectrum of Ship Machinery Vibration

Abstract

The present invention belongs to the field of ship vibration and noise control, and particularly relates to an intelligent identification method for characteristic line spectra of ship machinery vibration. The method comprises the following steps: classifying the characteristic data information of the ship machinery vibration characteristic line spectra to be identified; establishing a fitting numerical model of the power spectrum; establishing a continuous spectrum numerical model; dividing the vibration continuous spectrum into an upper part and a lower part with the continuous spectrum as the reference zero line; performing optimization of the local maximum of the line spectrum; The present application comprehensively considers several aspects such as the measurement point preprocessing of complex working condition data of multiple measurement points, the effective screening and extraction of the collected data, and the local search of the effective data sequence, so as to realize the rapid identification convergence and effective extraction of the characteristic line spectra in the power spectrum data of complex working conditions of multiple measurement points.

Description

Technical Field

[0001] The present invention belongs to the field of ship vibration and noise control, and particularly relates to an intelligent recognition method for characteristic line spectra of ship mechanical vibration. Background Art

[0002] Electromechanical equipment in a ship mechanical system exhibits significant vibration signals during operation, showing complex characteristic line spectra in the low-frequency band. These characteristic line spectra are transmitted to the water through the hull, forming radiated noise, which has an adverse impact on the acoustic stealth and comfort of the ship. The most common means of mechanical vibration monitoring is vibration signal analysis. The basic analysis methods of vibration signals are currently relatively mature, but the extraction methods for complex characteristic line spectra still need to be improved.

[0003] Currently, the methods applied to the extraction of characteristic line spectra of ship mechanical vibration are mainly based on spectrum estimation methods, which can be classified into: (1) noise envelope modulation detection or targeted wide-frequency segmented analysis and recording; (2) denoising the signal based on the wavelet threshold method and extracting the line spectra according to the singularity analysis results; (3) extracting characteristic line spectra according to ensemble empirical mode decomposition; (4) using an adaptive line spectrum enhancer to enhance the line spectra and suppress broadband interference during the line spectrum recognition and screening process; (5) manually identifying and extracting the line spectra.

[0004] The above means mainly focus on suppressing noise interference and improving the signal-to-noise ratio, making the line spectrum characteristics more obvious in the power spectrum signal. They have the disadvantages of large calculation errors, low operation efficiency, and limited placement of measurement points; some methods also require prior knowledge of the characteristic spectrum, increasing the difficulty of implementing the solution. Summary of the Invention

[0005] The purpose of the present invention is to provide an accurate and effective intelligent recognition method for characteristic line spectra of ship mechanical vibration on the premise of comprehensively considering the vibration coupling of measurement points, the complexity of measurement data, and the effectiveness of data extraction, so as to effectively extract the characteristic line spectra of ship mechanical vibration under complex working conditions.

[0006] The intelligent recognition method for characteristic line spectra of ship mechanical vibration of the present invention includes the following steps:

[0007] Step 1: Classify the characteristic line spectrum data information of the ship mechanical vibration to be recognized, including but not limited to: analyzing the measurement point name P, position number N, vibration signal preprocessing type F, data validity test set U, and vibration data preprocessing result set S; the above data sets constitute a database to be analyzed for vibration signals, mainly used to provide a data analysis basis for characteristic line spectra of vibration.

[0008] Step 2: Obtain the vibration power spectrum data from the data preprocessing result set in Step 1 and establish a fitting numerical model for the power spectrum. The power spectrum of mechanical vibration noise generally consists of a continuous spectrum and a line spectrum, which can be regarded as the superposition of the vibration characteristic line spectrum and the vibration continuous spectrum. Therefore, before extracting the characteristic line spectrum, a method for identifying the continuous spectrum should be provided to eliminate the continuous spectrum and obtain a relatively clean data base of the characteristic line spectrum flatness.

[0009] The power spectrum signal is usually a periodic signal, and its spectral characteristics are multiple single-line spectra. The line spectra are accompanied by main peaks and side lobes, and numerically, it shows that within the frequency band range near the characteristic line spectrum, its amplitude is the largest, and it decreases on both sides of the line spectrum.

[0010] Based on the above principle, the power spectrum sequence form is defined as:

[0011] [x1,x2,x3...x i ...x n-2 ,x n-1 ,x n

[0012] [y1,y2,y3...y i ...y n-2 ,y n-1 ,y n

[0013] In the above formula, x i is the frequency value obtained by fixed-step resolution processing, and y i is the corresponding power spectrum amplitude. Based on the wideband stationary random process model principle, the random vibration signal can be represented by the superposition of a stationary random signal and a phase random signal. Therefore, the continuous spectrum random vibration signal can be fitted as:

[0014]

[0015] In the above formula, {x(t)} is the wideband stationary random process function, and l r (t) represents a periodic signal with random phase. Further, the power spectrum can be expressed as:

[0016]

[0017] Among them, S is the spectral value, T is the time length of each signal segment participating in the operation when performing the primary Fourier transform (FFT), E is the arithmetic mean operator of the spectral value set, k is the signal segment number, and the value of K represents the signal segment accuracy.

[0018] Step 3: Establish a continuous spectrum numerical model based on the data base provided in Step 2. The vibration continuous spectrum can accurately reflect the change trend of the vibration amplitude-frequency characteristics within the frequency band. To balance the convergence speed and accuracy, a follow-up undetermined coefficient polynomial model is used to fit it, and its form can be expressed as:​​

[0019]

[0020] In the above formula, a is a follow-up undetermined coefficient, K is the fitting order, and m is the number of undetermined coefficients. Taking the minimum of the sum of the squares of the residuals as the fitting criterion, find the least-squares approximation curve τ of the undetermined coefficients:

[0021]

[0022] In the above formula, n is the number of power spectrum amplitude values in Step 1. To obtain the minimum value of τ, the undetermined coefficient a is treated one by one m Take the partial derivative:

[0023]

[0024] Let a m Set the partial derivative value to zero to obtain K + 1 linear equations:

[0025]

[0026] By solving the above equations, obtain the follow-up undetermined coefficient a of the fitting polynomial m and the fitting amplitude y that still contains the characteristic line spectrum * , and obtain the vibration continuous spectrum model from the undetermined coefficients.

[0027] Step 4: Using the vibration continuous spectrum as the reference zero line, divide the power spectrum amplitude values obtained in Step 2 into the upper-zero part and the lower-zero part according to whether the difference between the power spectrum and the continuous spectrum is greater than or less than 0.

[0028]

[0029] In the above formula, Y * and y * are the relative value of the power spectrum after removing the continuous spectrum and the fitting value of the continuous spectrum respectively. Perform a "0-1" judgment on the power spectrum amplitude sequence, and convert the corresponding frequency bands of the upper-zero part and the lower-zero part of Y * into multiple groups of continuous "1" intervals and continuous "0" intervals. The "0-1" judgment index equation for the effective characteristic line spectrum frequency values of each group is:

[0030] XOR(Y i-1 , Y i ) = 1, Y i = 1

[0031] XOR(Y i , Y i+1 ) = 1, Y i+1 = 0

[0032] In the above formula, when Y i-1 is below zero and Yi The frequency values above zero are set to 1, aiming to search for the starting value of the range where effective characteristic line spectra may appear, and for Y i above zero and Y i+1 The frequency values below zero are set to 0, aiming to search for the ending value of the range where effective characteristic line spectra may appear. Successive adjacent starting values and ending values form a group. Characteristic line spectra may appear within each group. Then, the corresponding power spectral amplitudes are extracted for each group to form a discrete data set with amplitudes as elements. Its discrete "0-1" grouping equation is:

[0033]

[0034] [0, Y1, Y2, …, Y n , 0], Y i ∈[0, 1], i = 1, 2…n

[0035] In the above formula, by judging the 0-1 relationship between the fitting value Y * and the preprocessed power spectral value set Y, the grouping of the power spectral amplitudes where characteristic line spectra may appear is realized. That is, the clustering of the effective data clusters of the power spectrum within the "1" interval of each group is achieved, thereby establishing a characteristic line spectrum data cluster that can perform local optimization. Compared with other methods, this step of establishing a local optimization data cluster of the characteristic line spectrum discards a large amount of invalid data, improving the recognition efficiency and the screening accuracy at the same time.

[0036] Step Five: Under the condition of meeting the 3dB signal-to-noise ratio, perform local maximum optimization of the line spectrum. Steps One to Four solve the problem of the possible frequency band range of the characteristic line spectrum. Utilizing the characteristics that a single characteristic line spectrum shows the largest line spectrum peak in the amplitude-frequency characteristic and the amplitudes on both sides of the peak decrease, sequentially extract the power spectral peak k peak in the k-th group of characteristic line spectrum data clusters, and respectively compare it with the mean value k e of each segment of the polynomial fitting curve obtained in Step Two. If k peak ≥ k e , it is judged as a valid line spectrum peak, and all valid peak sequences are extracted, and the number is counted as M; if k peak ≤ k e , the data is discarded. Finally, the effective line spectrum peaks are weighted and sorted or assigned according to the following four principles. The principles are in a parallel relationship:

[0037] (1) When the goal is to find the main characteristic line spectrum and the signal-to-noise ratio is strong (greater than 6dB), sort according to the absolute value of the peak;

[0038] (2) On the basis of (1), in order to extract line spectra with small absolute values but strong neighborhood contrast, sequential weight sorting needs to be performed according to the ratio of the mean values;

[0039] (3) To extract special line spectra with strong discreteness but low peaks, it is necessary to perform sequential weight sorting according to the variance of the corresponding power spectrum amplitudes.

[0040] (4) According to the vibration and sound characteristics and combined with the analysis of engineering requirements, the maximum weight can also be assigned to the line spectrum peaks within a certain frequency band range. After assigning weights according to the above four principles, the peak values and frequency sequences of the characteristic line spectra selected for the first time are finally identified. This step ensures that the main characteristic line spectra can be preferentially selected, and the characteristic line spectra with relatively high absolute values of adjacent amplitudes but relatively low absolute values of their own amplitudes will not be missed, and the special line spectra with strong discreteness caused by the acquisition step error will not be ignored.

[0041] Step 6: Check the obtained preliminary credible peak values and frequency sequences of the characteristic line spectra to judge the effective credibility of the data and detect possible missed peak values. Change the signal segment accuracy K in step 3, and take its highest power value within the range of [K - 5, K + 5] respectively. Repeat steps 3 to 5, and compare with the screening results of the characteristic line spectra based on the initial K - value vibration continuous spectrum. Take the difference set of each group of characteristic frequencies from the reference, and detect whether there are missed frequencies and peak values. If there are, take the screening results of adjacent unchanged K - values as the credible results.

[0042] The beneficial effects of the present invention are as follows:

[0043] The coupling relationship between the characteristic line spectra of each measuring point of the system reflects the vibration transmission path. Its intelligent recognition can achieve rapid extraction of the characteristic line spectra, which is of great significance for reducing the ship vibration noise level, accurately locating the vibration source, blocking the transmission path, and comprehensively improving the ship's stealth performance and comfort. The present invention realizes the effective recognition and directional screening of the characteristic line spectra of a large amount of mechanical vibration data under standard signal - to - noise ratio conditions through the vibration continuous spectrum quasi - zero baseline. Compared with traditional line spectrum extraction methods such as wavelet threshold method and adaptive line spectrum enhancer, it relies less on the prior knowledge of the characteristic spectrum, improves the recognition efficiency and screening accuracy of the characteristic line spectra, and reduces the difficulty of selecting characteristic line spectra in the large - scale ship mechanical vibration measurement project.

[0044] This application comprehensively considers several aspects such as the pre - processing of measuring points of multi - measuring - point complex working condition data, the effective screening and extraction of the collected data, and the local optimization of the effective data sequence, and realizes the rapid recognition convergence and effective extraction of the characteristic line spectra in the power spectrum data of multi - measuring - point complex working conditions. Description of the Drawings

[0045] Figure 1 It is the pre - processing power spectrum at the measuring point of the valve - controlled steering gear hydraulic motor foot.

[0046] Figure 2 It is the power spectrum at the measuring point of the foot and the corresponding continuous spectrum vibration continuous spectrum.

[0047] Figure 3 It is a characteristic line spectrum data cluster capable of local optimization;

[0048] Figure 4 It is the preliminarily extracted characteristic line spectrum;

[0049] Figure 5 It is a schematic diagram of the fitting accuracy test results under different K values;

[0050] Figure 6 It is a schematic diagram of the missed detection line spectrum result; Specific implementation manner

[0051] The present invention will be described in detail below in conjunction with specific implementation cases.

[0052] The structure of the ship power plant system is complex and there are many devices. When each ship is in different operating conditions, the independent operation and collaborative work of each device lead to complex vibrations in a wide frequency band, presenting irregular complex line spectrum characteristics in the test results, with strong coupling, randomness, complexity, and high requirements for signal-to-noise ratio. To realize the extraction of the characteristic line spectrum of the whole ship mechanical system, combined with the requirements of vibration source positioning and vibration intensity evaluation;

[0053] Taking the vibration test process of a marine valve-controlled steering gear as an example below, the intelligent identification method for the characteristic line spectrum of ship machinery vibration of the present application will be described in detail. Its basic steps include:

[0054] Step 1. As Figure 1 shown, preprocess the collected mechanical vibrations and establish a basic database of characteristic line spectra. Classify the characteristic line spectrum data information of the ship machinery vibration to be identified, aiming to facilitate the invocation of different time-domain or frequency-domain vibration signals by controlling variables. The classification includes but is not limited to: analyzing the measuring point name P, position number N, vibration signal preprocessing type F, data validity test set U, vibration data preprocessing result set S. The power spectrum analysis data in the subsequent steps uses the FFT preprocessing result.

[0055] Step 2. Establish a numerical model for fitting the test power spectrum and continuous spectrum of the marine steering gear, and call the vibration power spectrum preprocessing data in Step 1 to establish a numerical model for fitting the test power spectrum of the steering gear vibration at different measuring points respectively. Its power spectrum is regarded as composed of the superposition of the continuous spectrum and the line spectrum. The power spectrum sequences at different measuring points are all expressed in the following form:

[0056] [x1,x2,x3...x i ...x n-2 ,x n-1 ,x n P

[0057] [y1,y2,y3...y​i ...y n-2 ,y n-1 ,y n P

[0058] In the above formula, P is the measurement point number, x i is the frequency value obtained by processing with a fixed step resolution, y i is the corresponding power spectrum amplitude. Based on the principle of the wideband stationary random process model, the random vibration signals of each measurement point can be represented by the superposition of a stationary random signal and a phase random signal, and its continuous spectrum random vibration signal can be fitted as:

[0059]

[0060] In the above formula, {x(t)} is the wideband stationary random process function, l r (t) represents a periodic signal with random phase. Further, the vibration power spectrum of each measurement point of the servo can be expressed as:

[0061]

[0062] Among them, S is the spectral value, T is the time length of each signal segment participating in the operation when performing the primary Fourier transform (FFT), E is the arithmetic mean operator of the spectral value set, and k is the signal segment number.

[0063] Step 3: According to the servo power spectrum fitting numerical model provided in Step 2, establish a continuous spectrum numerical model for different measurement points. Its follow-up undetermined coefficient polynomial model is:

[0064]

[0065] In the above formula, a is the follow-up undetermined coefficient, K is the fitting order, and m is the number of undetermined coefficients. Taking the minimum of the sum of the squares of the residuals as the fitting criterion, find the least squares approximation curve τ of the undetermined coefficients:

[0066]

[0067] In the above formula, n is the number of power spectrum amplitudes defined in Step 1. Differentiate the undetermined coefficient a m one by one to obtain the minimum value of τ:

[0068]

[0069] Let a m The partial derivative value is zero, and then K + 1 linear equations are obtained:

[0070]

[0071] By solving the above equations, the follow-up undetermined coefficients a of the fitting polynomial are obtained​m and the fitted amplitude y that still contains the characteristic line spectrum * , obtain the vibration continuous spectrum model from the undetermined coefficients, and the results are as Figure 2 shown.

[0072] Step 4: Using the continuous spectrum (vibration continuous spectrum) as the reference zero line, divide the power spectrum amplitude obtained from the preprocessing into the upper part and the lower part of zero:

[0073]

[0074] In the above formula, Y * and y * are the relative value of the power spectrum after removing the continuous spectrum and the fitted value of the continuous spectrum respectively. Through the judgment of the power spectrum amplitude sequence "0-1", the corresponding frequency bands of the upper part and the lower part of Y* are respectively converted into multiple groups of continuous "1" intervals and continuous "0" intervals. Among them, the index equations of the effective characteristic line spectrum frequency values of each group are:

[0075] XOR(Y i-1 , Y i ) = 1, Y i = 1

[0076] XOR(Y i , Y i+1 ) = 1, Y i+1 = 0

[0077] In the above formula, the frequency value where Y i-1 is below zero and Y i is above zero is set to 1, aiming to search for the starting value of the range where effective characteristic line spectra may appear. The frequency value where Y i is above zero and Y i+1 is below zero is set to 0, aiming to search for the ending value of the range where effective characteristic line spectra may appear, so as to realize the frequency band grouping of the possible characteristic line spectra. Then, the power spectrum amplitudes corresponding to the possible characteristic line spectrum frequency band grouping are grouped separately to form discrete data, and its discrete "0-1" grouping equation is:

[0078]

[0079] [0, Y1, Y2, …, Y n , 0], Y i ∈[0, 1], i = 1, 2…n

[0080] In the above formula, by judging the fitted value Y *The 0-1 relationship with the set Y of preprocessed power spectrum values is used to group the power spectrum amplitude values of the possible characteristic line spectra. That is, the clustering of the effective data clusters of the power spectrum within the "1" interval of each group is realized, so as to establish a characteristic line spectrum data cluster that can be locally optimized. Taking the measuring point of the machine foot as an example, the obtained characteristic line spectrum data cluster is as Figure 3 shown.

[0081] Step 5: Optimize the local maximum of the line spectrum under the condition of satisfying the 3 dB signal-to-noise ratio. After the "0-1" recognition and screening in Step 4, the power spectrum fitting model formed in Step 2 is also divided into a fitting numerical model segmented by frequency. The positive part of the continuous power spectrum data of each measuring point with a fixed step size is converted into discontinuous data, and further encoded into k groups of continuous characteristic line spectrum data clusters through an indexing logical formula.

[0082] For each measuring point, sequentially extract the line spectrum peak value k in the kth group of continuous characteristic line spectrum data clusters peak , and respectively compare it with the mean value k of each section of the continuous spectrum (vibration continuous spectrum) fitting numerical model obtained in Step 2 e . If k peak ≥k e , it is judged as an effective peak value. If k peak ≤k e , the data is discarded. Form M effective peak value sequences, and perform weight sorting or assignment according to the following four principles: (1) Sort in descending order according to the absolute value of the power spectrum peak; (2) Take the first characteristic line spectrum as the reference value and perform sequential weight sorting according to the ratio; (3) According to the vibration and sound characteristics, assign the maximum weight to the line spectrum peak value in the range of 10 - 1000 Hz according to the implementation plan; (4) Perform sequential weight sorting according to the variance of the corresponding power spectrum amplitude. Finally, the characteristic line spectrum peak value and frequency sequence are obtained, which can be used as the basis for extracting the characteristic line spectrum.

[0083] Step 6: Check the extraction results. Change the K value to change the signal segment accuracy. In the range of [K - 5, L + 5], take the existence of the minimum value of the least squares approximation curve τ in Step 3 as the undetermined coefficient a mBoundary conditions for value taking, respectively take their highest power values, repeat steps 2 - 5, and compare them with the screening results of the K value selected for the first time (the initially selected K is 5). For each set of characteristic frequencies, take the difference set with the benchmark to detect whether there are missed detection frequencies. If there are, when the screening results of K - 1, K, and K + 1 are unchanged, it is a credible missed detection frequency; otherwise, it is not credible. For example, select the initial K value as 5. Through the least - squares trial calculation in step three, it is determined that K should be greater than or equal to 5. Therefore, select K values as [5, 6, 7, 8, 9, 10] respectively. The number of characteristic line spectra in the power spectrum (0 - 1000Hz) is 19, 19, 18, 18, 18, 18 respectively, and the repetition rate of the corresponding identified characteristic frequencies is relatively high. It can be seen from Figure 5 that the highest power is not the higher the better. When K = 5, since the vibration continuous spectrum changes more flatly than higher powers, 809Hz is missed because it is too close to the 814Hz characteristic line spectrum nearby and is divided into the same group; when K = 6 - 10, 465Hz and 482Hz are grouped together because the fitting value of the vibration continuous spectrum is distorted here due to the relatively high highest power. Therefore, the highest power of polynomial fitting can be taken as 5.

[0084] After merging the characteristic line spectra at the five sets of servo foot measurement points, a total of 20 overlapping frequencies are removed. Taking the screening result with K = 5 as the comparison benchmark, take the difference set of each set of characteristic frequencies with it to detect the missed detection frequencies, as shown in Table 1 and Figure 5 as shown.

[0085] It can be seen that the power spectrum and its continuous spectrum (vibration continuous spectrum) obtained by using the inventive method are as Figure 2 shown; the data cluster of characteristic line spectra that can be used for local optimization obtained by judging through the zero line is as Figure 3 shown; when K = 5, the peak sequence of the characteristic line spectrum finally extracted according to the inventive method is as Figure 4 shown, and the final test result is as Figure 5 、 Figure 6 shown. Through the test, the data results of line spectrum extraction given in Table 1 are extracted.

[0086] Table 1 Characteristic line spectrum extraction results

[0087]

[0088] By comparing with Table 1 in step six, it can be seen that the accuracy of the characteristic line spectrum extracted by the method is relatively high. Only the dense small characteristic line spectra near individual main characteristic line spectra are not detected, but the missed detection is corrected in step six, verifying the reliability of the method.

[0089] Finally, it should be noted that the above cases are only used to illustrate the technical solutions of the present invention, rather than the development of the protection scope of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A rapid identification and screening method for the characteristic line spectrum of ship machinery vibration, characterized in that, It includes the following steps: Step 1: A step for constructing a database to be analyzed of vibration signals; Specifically, it includes: classifying the characteristics of the line spectrum data of the ship machinery vibration to be identified, forming a characteristic classification data set, and composing the data set into a database to be analyzed of vibration signals; Step 2: A step for establishing a numerical model for power spectrum fitting Specifically, it includes: obtaining vibration power spectrum data from the database to be analyzed and establishing a numerical model for power spectrum fitting: In the above formula, is the frequency value obtained by processing with a fixed step resolution, is the corresponding power spectrum amplitude; Based on the principle of the wideband stationary random process model, the random vibration signal is represented by the superposition of a stationary random signal and a phase random signal, and the continuous spectrum random vibration signal is fitted as: In the above formula, is a wideband stationary random process function, is a periodic signal with random phase; Furthermore, the power spectrum is expressed as: Among them, S is the spectral value, T is the time length of each signal segment participating in the operation during the primary Fourier transform, E is the arithmetic mean operator of the spectral value set, and k is the signal segment number; Step 3: A step for establishing a numerical model of the continuous spectrum Specifically, it includes: establishing a numerical model of the continuous spectrum according to the flat data base of the characteristic line spectrum; the vibration continuous spectrum reflects the change trend of the vibration amplitude-frequency characteristic within the frequency band, and it is fitted by using a follow-up undetermined coefficient polynomial model, and its form is expressed as: In the above formula, is an undetermined coefficient, K is the fitting order, and m is the number of undetermined coefficients; the value of K represents the signal segment accuracy; Taking the minimum of the sum of the squares of the residuals as the fitting criterion, finding the least squares approximation curve τ of the undetermined coefficients: In the above formula, n is the number of power spectrum amplitudes in the first step. To obtain the minimum value of τ, each undetermined coefficient is treated one by one Take the partial derivative: , where ; Let the partial derivative value be zero, and we obtain a system of linear equations: , where ; ; By solving the above system of equations, the undetermined coefficients of the fitting polynomial are obtained and the fitting amplitude still containing the characteristic line spectrum , and a vibration continuous spectrum model is obtained from the undetermined coefficients; Step 4: A step for establishing a cluster of characteristic line spectrum data that can perform local optimization Taking the vibration continuous spectrum as the quasi-zero line, dividing the amplitude of the power spectrum obtained by preprocessing into the upper part and the lower part according to whether the difference between the power spectrum and the continuous spectrum is greater than or less than 0; In the above formula, is the relative value of the power spectrum after removing the continuous spectrum; is the fitting value of the continuous spectrum; Judging by the power spectrum amplitude sequence 0-1, respectively convert the corresponding frequency bands of the above-zero part and the below-zero part of into multiple groups of continuous 1 intervals and continuous 0 intervals; obtain the index equations of the effective characteristic line spectrum frequency values for each group: , In the above formula, The frequency values below zero and above zero are set to 1, aiming to search for the starting value of the range where effective characteristic line spectra may appear; The frequency values above zero and below zero are set to 0, aiming to search for the ending value of the range where effective characteristic line spectra may appear, and to achieve the frequency band grouping of the possible characteristic line spectra; then the power spectrum amplitudes corresponding to each group of the possible characteristic line spectrum frequency bands are grouped separately to form discrete data, and its discrete 0-1 grouping equation is: In the above formula, by judging the 0-1 relationship between the fitting value and the set Y of preprocessed power spectrum values, the power spectrum amplitude grouping of the possible characteristic line spectra is realized; the clustering of the effective data clusters of the power spectrum in the 1 interval of each group is realized, so as to establish a characteristic line spectrum data cluster that can perform local optimization; Step 5: A step for the weight ranking of the peak value and frequency sequence of the characteristic line spectrum; First, under the condition of satisfying the 6 dB signal-to-noise ratio, optimize the local maximum value of the line spectrum. Secondly, utilize the characteristics that a single characteristic line spectrum shows the maximum peak value in the amplitude-frequency characteristic and the amplitudes on both sides of the peak value decrease. Sequentially extract the power spectrum peak values in the k-th group of characteristic line spectrum data clusters , and respectively compare them with the mean values of the polynomial fitting curves obtained in the second step ; if , then it is judged as a valid line spectrum peak value, and extract all valid peak value sequences, and the number is counted as M; finally, perform weight sorting or assignment on the characteristic line spectrum peak values to identify the initially credible characteristic line spectrum peak values and frequency sequences Step 6: Testing the obtained preliminary credible peak value and frequency sequence results of the characteristic line spectrum to judge the effective information of the data and detecting the possible missed detected peaks; Changing the signal segment accuracy K in Step 3, respectively taking its highest power value within the range of [K - 5, K + 5], repeating Steps 3 to 5, and comparing with the screening results of the characteristic line spectrum with the initial K value vibration continuous spectrum as the identification basis, making a difference set of each group of characteristic frequencies and the reference, detecting whether there are missed detected frequencies and peaks, if any, taking the screening results of the adjacent unchanged K value as the credible result.

2. The rapid identification and screening method for the characteristic line spectrum of ship machinery vibration according to claim 1, characterized in that, In the said Step 1, the said characteristic classification includes: analyzing the measuring point name P, position number N, vibration signal preprocessing type F, data validity test U, and vibration data preprocessing result S.

3. A method for quickly identifying and screening the characteristic line spectrum of ship machinery vibration according to claim 1, characterized in that, In the said Step 2, the power spectrum of the mechanical vibration is regarded as the superposition on the vibration characteristic line spectrum and the vibration continuous spectrum, and the continuous spectrum should be eliminated before extracting the characteristic line spectrum.

4. A method for quickly identifying and screening the characteristic line spectrum of ship machinery vibration according to claim 1, characterized in that In the said Step 5, the weight ranking or assignment is carried out according to the following four orders to identify the preliminary credible peak value and frequency sequence of the characteristic line spectrum; (1) Ranking according to the absolute value of the peak value; (2) Sequential weight ranking according to the ratio size; (3) According to the vibration-acoustic characteristics, giving the maximum weight to the peak value of the line spectrum within a partial frequency band range; (4) Sequential weight ranking according to the variance of the corresponding power spectrum amplitude; Finally, the preliminary credible peak value and frequency sequence of the characteristic line spectrum are identified.

Images