Marine main engine reciprocating friction vibration signal distinguishing method based on multi-scale permutation entropy
By adaptively selecting IMF components based on correlation coefficients and using a multi-scale permutation entropy algorithm, combined with a machine learning classifier, the problem of inaccurate signal selection after CEEMD decomposition was solved, and efficient monitoring of reciprocating friction vibration signals of ship main engines was achieved.
Patent Information
- Application Number
- CN202511514273.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-22
- Publication Date
- 2026-02-03
AI Technical Summary
In existing technologies, CEEMD decomposition cannot effectively filter out effective signals and noise signals from the reciprocating friction vibration signals of ship main engines, resulting in reduced monitoring accuracy.
An adaptive correlation coefficient method was used to screen IMF components, and a multi-scale permutation entropy algorithm was used to identify vibration signals, which were then combined with a machine learning classifier for recognition.
It improves the monitoring accuracy of reciprocating friction vibration signals of ship main engines, and can quickly and effectively filter out noise, thus improving judgment efficiency.
Smart Images
Figure CN121456584A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data / signal processing methods, and in particular to a method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale permutation entropy. Background Technology
[0002] Monitoring and maintaining the condition of ship main engines has always been a significant challenge. Monitoring the reciprocating friction vibration signals of ship main engines plays a crucial role in this process, as these signals contain vital information about the engine's operational status. However, these signals are often subject to noise interference and may contain vibration components from multiple sources, complicating the accurate identification of abnormal conditions.
[0003] In existing technologies, the EEMD method has been further improved by incorporating auxiliary noise using a positive and negative pairwise method to eliminate residual auxiliary noise in the reconstructed signal. Simultaneously, the number of noise sets added can be minimized, resulting in high computational efficiency. This method is known as Complementary Set Empirical Mode Decomposition (CEEMD). However, in practical applications, existing screening methods, such as direct spectrum screening, correlation coefficient screening, kurtosis screening, and adaptive noise reduction, fail to effectively and correctly synthesize effective and noise signals from the extracted IMF components after CEEMD decomposition, leading to a decrease in the accuracy of monitoring ship main engine reciprocating friction vibration signals.
[0004] Therefore, this application provides a method for screening and reconstructing IMFs using an adaptive correlation coefficient method, and for identifying vibration signals under different frictional states using a multi-scale permutation entropy algorithm. Summary of the Invention
[0005] To address the aforementioned problems in the existing technology, the purpose of this invention is to provide a method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale permutation entropy.
[0006] Another objective of this invention is to provide a system for identifying reciprocating friction vibration signals of a ship's main engine based on multi-scale permutation entropy.
[0007] To solve the above problems, the present invention adopts the following technical solution: a method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale permutation entropy, the method comprising the following steps:
[0008] Step 1: Collect reciprocating friction vibration signals of the ship's main engine piston;
[0009] Step 2: Use CEEMD to decompose the acquired vibration signal and obtain several IMF components and the last residual quantity;
[0010] Step 3: Reconstruct the signal from the decomposed IMF components using the correlation coefficient adaptive method;
[0011] Step 4: Compare the reconstructed signal with the experimental signal using multi-scale permutation entropy to determine the partial mean of the multi-scale permutation entropy;
[0012] Step 5: Use the multi-scale permutation entropy partial mean as the feature vector, input the feature vector into a machine learning-based classifier to identify the vibration signal and determine the friction state corresponding to the signal.
[0013] Furthermore, as described in step two, the CEEMD decomposition signal algorithm function specifically employs a positive-negative pairing method to eliminate residual auxiliary noise in the reconstructed signal by applying positive and negative pairs to the auxiliary noise added by the EEMD method.
[0014] Furthermore, the correlation coefficient adaptive algorithm described in step three is specifically as follows:
[0015]
[0016] Where CC is the correlation coefficient between the IMF component and the original signal, r represents the IMF component, x represents the vibration signal, n represents the number of IMF components, and i represents the index of the i-th data point; CC represents the degree of correlation between signals. The larger the CC value, the greater the correlation between r and x; if CC is close to 0, it indicates that the correlation between r and x is weak; by setting a threshold θ, the IMF component with a larger correlation coefficient is selected. .
[0017] Furthermore, in step four, the permutation entropy is first calculated; considering a time series of length N {x(i), i=1,2,⋯,N}, its phase space is reconstructed to obtain the following time series:
[0018] In the formula, m is the embedding dimension; λ is the time delay;
[0019] Rearrange the m data points of X(i) in ascending order, i.e. .
[0020] If it exists Sort by the size of the j value, that is, when < ,have For any data X(i), a sequence of symbols can be obtained:
[0021] in, m different symbols There are m! distinct permutations and m! distinct symbol sequences, where S(g) is one of the m! symbol sequences; calculate the probability of each symbol sequence occurring. At this point, the time series The permutation entropy is defined in the form of Shannon entropy as
[0022] right standardization, Right now, The range of values for is 0≤ ≤1;
[0023] Multi-scale permutation entropy is defined as the permutation entropy at different scales, and its calculation method is as follows:
[0024] Consider the time series {x(i), i=1,2, The sequence { ,N} is coarsened to obtain a coarse-grained sequence { ; The expression is
[0025]
[0026] in, express Rounding down; As a scale factor, =1,2, , =1, the coarse-grained sequence is the original sequence; When the length is greater than 1, the original sequence is coarsened to a length of [length missing]. The coarse-grained sequences are calculated; for each coarse-grained sequence, its permutation entropy value is calculated, and the results are plotted as permutation entropy functions at different scales. A scale factor greater than 10 is chosen to ensure more effective noise smoothing, highlighting long-term trends and periodic patterns, and improving the stability of statistical results when analyzing time series.
[0027] Furthermore, as shown in step five, the partial mean of the multi-scale permutation entropy is used as the feature vector, specifically by calculating the skewness S of the multi-scale permutation entropy. ke , that is, the ratio of the absolute value of the skewness of the sequence to its standard deviation, is calculated using the following formula:
[0028]
[0029] In the formula, , and Representing the entropy of multi-scale permutations respectively The mean, median, and standard deviation;
[0030] The partial mean of the multi-scale arrangement entropy of the vibration signal is obtained by the following formula.
[0031] .
[0032] This application also provides a system for the aforementioned method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale permutation entropy, the system comprising:
[0033] Acquisition unit: configured to acquire vibration signals of the ship's main engine under different vibration and friction conditions;
[0034] The signal processing unit is configured to decompose the acquired vibration signal using CEEMD, obtain the decomposed intrinsic mode functions (IMFs), reconstruct the signal using an adaptive correlation coefficient method for the decomposed IMF components, and compare the reconstructed signal with the experimentally obtained signal to determine the partial mean of the multi-scale permutation entropy.
[0035] The discrimination unit is configured to use the partial mean of multi-scale entropy as a feature vector, input the feature vector into a machine learning-based classifier, identify the vibration signal, and distinguish the friction state corresponding to the signal.
[0036] This application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps in the method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale permutation entropy.
[0037] This application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the method for identifying reciprocating friction vibration signals of a ship's main engine based on multi-scale permutation entropy.
[0038] Compared with the prior art, the beneficial technical effects of the present invention are as follows:
[0039] 1. The method described in this invention, by providing a method for adaptively filtering IMF components based on correlation coefficients and then reconstructing the IMF components, compared with existing methods such as direct spectrum filtering, correlation coefficient filtering, kurtosis filtering, and adaptive noise reduction, quickly and effectively filters out noise from the original signal after vibration signal measurement, rather than filtering through manually set parameters. Compared with judging the friction state by comparing it with normal three-dimensional morphology, it can directly determine the friction state of the measured signal by comparing the multi-scale arrangement entropy of the signal, thus improving the efficiency of judgment. Attached Figure Description
[0040] Figure 1 This is a flowchart illustrating an embodiment of this application;
[0041] Figure 2 This is a time-domain waveform diagram of a test signal collected in an embodiment of this application;
[0042] Figure 3 This is a waveform diagram of the environmental noise signal collected in an embodiment of this application;
[0043] Figure 4 This is a CEEMD decomposition residual white noise diagram of an embodiment of this application;
[0044] Figure 5 This is a waveform diagram of the clean signal obtained after signal reconstruction in an embodiment of this application;
[0045] Figure 6 This is a waveform diagram of the noise signal obtained after signal reconstruction in an embodiment of this application;
[0046] Figure 7 This is a graph showing the entropy curve of a pure signal arranged in multiple scales according to an embodiment of this application.
[0047] Figure 8 This is a multi-scale permutation entropy curve of the noise signal in an embodiment of this application;
[0048] Figure 9 This is a graph showing the partial mean of the entropy of the pure signal and the multi-scale permutation obtained by the laboratory algorithm in the embodiments of this application.
[0049] Figure 10 This is a multi-scale entropy partial mean curve of noise signal and environmental noise signal in an embodiment of this application. Detailed Implementation
[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0051] Example 1
[0052] like Figure 1 As shown, a method for identifying reciprocating frictional vibration signals of ship main engines based on multi-scale permutation entropy is described, the method comprising the following steps:
[0053] Step 1: Collect reciprocating friction vibration signals of the ship's main engine piston;
[0054] Step 2: Use CEEMD to decompose the acquired vibration signal to obtain several IMF components and a residual quantity. The CEEMD decomposition algorithm function specifically uses a positive and negative pairing method to eliminate the residual auxiliary noise in the reconstructed signal.
[0055] Step 3: Reconstruct the signal from the decomposed IMF components using an adaptive correlation coefficient method; the adaptive correlation coefficient algorithm is specifically as follows:
[0056]
[0057] Where CC is the correlation coefficient between the IMF component and the original signal, r represents the IMF component, x represents the vibration signal, n represents the number of IMF components, and i represents the index of the i-th data point; CC represents the degree of correlation between signals. The larger the CC value, the greater the correlation between r and x; if CC is close to 0, it indicates that the correlation between r and x is weak; by setting a threshold θ, the IMF component with a larger correlation coefficient is selected.
[0058] .
[0059] Step 4: Compare the reconstructed signal with the experimental signal using multi-scale permutation entropy to determine the partial mean of the multi-scale permutation entropy;
[0060] Step 5: Using the partial mean of the multi-scale arrangement entropy as a feature vector, input the feature vector into a machine learning-based classifier to identify the vibration signal and determine the corresponding friction state. Specifically, the partial mean of the multi-scale arrangement entropy is used as the feature vector, and the skewness S of the multi-scale arrangement entropy is calculated. ke , that is, the ratio of the absolute value of the skewness of the sequence to its standard deviation, is calculated using the following formula:
[0061]
[0062] In the formula, , and Representing the entropy of multi-scale permutations respectively The mean, median, and standard deviation;
[0063] The partial mean of the multi-scale arrangement entropy of the vibration signal is obtained by the following formula.
[0064] .
[0065] The specific implementation includes the following steps:
[0066] Step 1: First, the sensor collects multiple segments of vibration signals from the target ship's main engine. The collected signals are then input into the data processing unit and into the MATLAB program for the programmed algorithm. In this embodiment, one segment of the test signal collected by the sensor is shown below. Figure 2 As shown. And the environmental noise signal collected by the sensor is as follows: Figure 3 As shown.
[0067] Step 2: After acquiring the signal, the original signal is decomposed using the CEEMD signal decomposition method, yielding several IMF components and a residual signal. This residual signal is used to prepare for reconstructing the effective and noise signals from the subsequent IMF components. The residual white noise obtained through CEEMD decomposition is shown below. Figure 4 As shown
[0068] Step 3: Calculate the correlation coefficient threshold of the IMF component in each signal segment to determine the range of the reconstructed effective signal and noise signal. The adaptive correlation coefficient algorithm is as follows: Where CC is the correlation coefficient between the IMF component and the original signal, r is the IMF component, x represents the vibration signal, and n is the number of IMF components. CC indicates the degree of correlation between the signals; the larger the CC value, the stronger the correlation between r and x. If CC is close to 0, it indicates that the correlation between r and x is weak. By setting a threshold θ, the IMF component with a larger correlation coefficient is selected. The signal is reconstructed based on the adaptively calculated threshold. The threshold for the correlation coefficient is calculated to be 0.24541. The Pearson correlation coefficient between each IMF and the vibration signal is shown in the table below.
[0069] surface 1. Pearson correlation coefficients of IMFs obtained from CEEMD decomposition of the original signal
[0070] The clean signal and noise signal obtained after signal reconstruction are respectively as follows: Figure 5 and Figure 6 As shown
[0071] The obtained original signals are all subjected to the above-mentioned CEEMD decomposition and reconstruction.
[0072] Step 4: Calculate the multi-scale permutation entropy of the extracted multi-segment signals, and calculate its partial mean. Calculate the skewness S of the multi-scale permutation entropy. ke , which is the ratio of the absolute value of the skewness of the sequence to its standard deviation, is calculated using the following formula: In the formula, , and Representing the entropy of multi-scale permutations respectively The mean, median, and standard deviation of the vibration signal. The partial mean of the multi-scale arrangement entropy of the vibration signal can be obtained by the following formula.
[0073] .
[0074] The multi-scale entropy curves of vibration signals with different friction types have different rates of descent, and the rate at which the complexity of vibration signals under different friction conditions decreases with increasing scale is also different.
[0075] Step 5: To better distinguish the different operating states of reciprocating friction vibration signals, the partial mean of multi-scale arrangement entropy is used as its characteristic index. The multi-scale arrangement entropy and the partial mean of multi-scale arrangement entropy in the embodiment are calculated. The multi-scale arrangement entropy of the pure signal is as follows: Figure 7 As shown, the multi-scale permutation entropy of the noise signal is as follows: Figure 8 As shown, the partial mean of the entropy of the pure signal and the multi-scale permutation obtained by the laboratory algorithm is as follows: Figure 9 As shown, the partial mean of the multi-scale permutation entropy of the noise signal and the environmental noise signal is as follows: Figure 10 As shown.
[0076] Example 2
[0077] A system for identifying reciprocating friction vibration signals of ship main engines based on multi-scale permutation entropy, the system comprising:
[0078] Acquisition unit: configured to acquire vibration signals of the ship's main engine under different vibration and friction conditions;
[0079] The signal processing unit is configured to decompose the acquired vibration signal using CEEMD, obtain the decomposed intrinsic mode functions (IMFs), reconstruct the signal using an adaptive correlation coefficient method for the decomposed IMF components, and compare the reconstructed signal with the experimentally obtained signal to determine the partial mean of the multi-scale permutation entropy.
[0080] The discrimination unit is configured to use the partial mean of multi-scale entropy as a feature vector, input the feature vector into a machine learning-based classifier, identify the vibration signal, and distinguish the friction state corresponding to the signal.
[0081] Example 3
[0082] This application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps in the method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale permutation entropy.
[0083] Example 4
[0084] This application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the method for identifying reciprocating friction vibration signals of a ship's main engine based on multi-scale permutation entropy.
[0085] Finally, it should be pointed out that the above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for identifying reciprocating frictional vibration signals of ship main engines based on multi-scale permutation entropy, characterized in that, The method includes the following steps: Step 1: Collect reciprocating friction vibration signals of the ship's main engine piston; Step 2: Use CEEMD to decompose the acquired vibration signal and obtain several IMF components and a residual quantity; Step 3: Reconstruct the signal from the decomposed IMF components using the correlation coefficient adaptive method; Step 4: Compare the reconstructed signal with the experimental signal using multi-scale permutation entropy to determine the partial mean of the multi-scale permutation entropy; Step 5: Use the multi-scale permutation entropy partial mean as the feature vector, input the feature vector into a machine learning-based classifier to identify the vibration signal and determine the friction state corresponding to the signal.
2. The method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale arrangement entropy according to claim 1, characterized in that, As described in step two, the CEEMD decomposition signal algorithm function specifically uses a positive and negative pairing method to eliminate residual auxiliary noise in the reconstructed signal by applying positive and negative pairs to the auxiliary noise added by the EEMD method.
3. The method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale arrangement entropy according to claim 1, characterized in that, The correlation coefficient adaptive algorithm described in step three is as follows: ， Where CC is the correlation coefficient between the IMF component and the original signal, r represents the IMF component, x represents the vibration signal, n represents the number of IMF components, and i represents the index of the i-th data point; CC represents the degree of correlation between signals. The larger the CC value, the greater the correlation between r and x; if CC is close to 0, it indicates that the correlation between r and x is weak; by setting a threshold θ, the IMF component with a larger correlation coefficient is selected. .
4. The method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale arrangement entropy according to claim 1, characterized in that, In step four, the permutation entropy is first calculated. A time series of length N {x(i), i=1,2,⋯,N} is reconstructed in phase space to obtain the following time series: , In the formula, m is the embedding dimension; λ is the time delay; Rearrange the m data points of X(i) in ascending order, i.e. ; If it exists Sort by the size of the j value, that is, when < ,have For any data X(i), a sequence of symbols can be obtained: in, m different symbols There are m! different permutations, and correspondingly, m! different symbol sequences, where S(g) is one of the m! symbol sequences; calculate the probability of each symbol sequence occurring. Time series The permutation entropy is defined in the form of Shannon entropy as , right standardization, Right now, The range of values for is 0≤ ≤1; Multi-scale permutation entropy is defined as the permutation entropy at different scales, and is calculated as follows: Consider the time series {x(i), i=1,2, The sequence { ,N} is coarsened to obtain a coarse-grained sequence { ; The expression is ， in, express Rounding down; As a scale factor, =1,2, ; =1, the coarse-grained sequence is the original sequence; When the length is greater than 1, the original sequence is coarsened to a length of [length missing]. The coarse-grained sequences are calculated; for each coarse-grained sequence, its permutation entropy value is calculated, and the results are plotted as permutation entropy functions at different scales.
5. The method for identifying reciprocating friction vibration signals of ship main engines based on multi-scale arrangement entropy according to claim 1, characterized in that, As shown in step five, the partial mean of the multi-scale permutation entropy is used as the feature vector. Specifically, the skewness S of the multi-scale permutation entropy is calculated. ke The ratio of the absolute value of the skewness of a sequence to its standard deviation is calculated using the following formula: ， In the formula, , and Representing the entropy of multi-scale permutations respectively The mean, median, and standard deviation; The partial mean of the multi-scale arrangement entropy of the vibration signal is obtained by the following formula. 。 6. A system for implementing the method for identifying reciprocating friction vibration signals of a ship's main engine based on multi-scale permutation entropy as described in claim 1, characterized in that, The system includes: Acquisition unit: configured to acquire vibration signals of the ship's main engine under different vibration and friction conditions; The signal processing unit is configured to decompose the acquired vibration signal using CEEMD, obtain the decomposed intrinsic mode functions (IMFs), reconstruct the signal using an adaptive correlation coefficient method for the decomposed IMF components, and compare the reconstructed signal with the experimentally obtained signal to determine the partial mean of the multi-scale permutation entropy. The discrimination unit is configured to use the partial mean of multi-scale entropy as a feature vector, input the feature vector into a machine learning-based classifier, identify the vibration signal, and distinguish the friction state corresponding to the signal.
7. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the steps in the method for identifying reciprocating friction vibration signals of a ship's main engine based on multi-scale permutation entropy, as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the method for identifying reciprocating friction vibration signals of a ship's main engine based on multi-scale permutation entropy, as described in any one of claims 1 to 5.