A multi-parameter optimized internal combustion engine noise source separation method

Abstract

This invention provides a multi-parameter optimized method for separating noise sources in an internal combustion engine, comprising: collecting acoustic and vibration signals of the internal combustion engine under typical operating conditions and reversing conditions; preprocessing the collected acoustic signals and using them as inputs for variational mode decomposition (VMD); using the number of VMD modes and penalty factor as optimization variables; employing an improved AO optimization algorithm combined with an energy-weighted fuzzy entropy fitness function to adaptively optimize VMD parameters and perform signal decomposition to obtain several IMF components; combining time-frequency analysis, coherence analysis, and prior knowledge to separate and identify non-order noise, periodic impact noise, and order noise, thereby completing noise source separation; wherein, the improvements to the improved AO optimization algorithm include: introducing a dynamic disturbance factor and an adaptive probability coefficient; the energy-weighted fuzzy entropy fitness function calculates a weighted average entropy value as a fitness index by combining the energy distribution of the IMF components with the fuzzy entropy measurement.

Description

Technical Field

[0001] This invention relates to the field of internal combustion engine noise separation technology, and more specifically to a multi-parameter optimized method for separating internal combustion engine noise sources. Background Technology

[0002] With the development of the internal combustion engine industry, vehicle NVH performance has become one of the core indicators for measuring product competitiveness, and the requirements for engine comfort are becoming increasingly stringent. The noise generated by an internal combustion engine during operation is complex, mainly including combustion noise, piston knocking noise, gear meshing noise, and fluid noise. Accurately separating these noise sources is key to fault diagnosis and noise control.

[0003] Early noise source separation largely relied on traditional signal processing methods. In 1998, Huang et al. proposed Empirical Mode Decomposition (EMD), which adaptively decomposes complex signals into a series of intrinsic mode functions (IMFs) based on the time scale of discrete data, offering certain advantages in intermittent signal processing (Huang NE, Shen Z, Long SR, et al. The Empirical Mode Decomposition and the Hilbert Spectrum for Nonlinear and Non-Stationary TimeSeries Analysis [J]. Proceedings Mathematical Physical & Engineering Sciences, 1998, 454(1971): 903-995.). However, the EMD algorithm still has significant limitations in complex engineering applications, such as the susceptibility to mode aliasing, endpoint effects, energy leakage, under-decomposition, and over-decomposition. Furthermore, it is sensitive to noise, easily introducing random noise components into the effective modes, thus reducing the decomposition quality.

[0004] To address the shortcomings of EMD, Dragomiretskiy et al. proposed the Variational Mode Decomposition (VMD) algorithm in 2014 (Dragomiretskiy K, Zosso D. Variational Mode Decomposition [J]. IEEE Transactions on Signal Processing, 2014, 62(3): 531-544.). This method uses a variational decomposition mode instead of the traditional recursive decomposition, which has a more solid mathematical theoretical foundation and the signal decomposition results have better accuracy and robustness, effectively avoiding the inherent defects of EMD and its derivative algorithms. However, VMD requires manual setting of parameters before signal decomposition. The quality of parameter selection directly determines the decomposition effect, and human factors make it difficult to guarantee the accuracy of the decomposition results.

[0005] Li Z et al. optimized VMD, but only optimized the number of decomposed modes K, without considering the influence of the penalty factor α (Li Z, Chen J, Zi Y, et al. Independence-oriented VMD to identify faultfeature for wheel set bearing fault diagnosis of high speed locomotive [J].Mechanical systems and signal processing, 2017, 85: 512-529.). Shi P et al. used optimized VMD to study feature extraction and condition monitoring of wind turbines. This study optimized the number of decomposed modes K and the penalty factor α separately, but ignored the coupling effect of the two parameters (Shi P, Yang W. Precise feature extraction from wind turbine condition monitoring signals by using optimized variational mode decomposition [J]. IET Renewable Power Generation, 2017, 11(3): 245-252.). Xiao H et al. proposed a VMD parameter optimization method based on Fourier spectral entropy as the fitness function to identify the low-frequency features of the system. This method considers the coupling effect of K and α and optimizes these two key parameters. However, Fourier spectral entropy is not suitable for non-steady-state signal processing such as diesel engines (Xiao H, Wei J and Liu H, et al. Identification method for power system low-frequency oscillations based on improved VMD and Teager-Kaiser energy operator [J]. IET Generation, Transmission & Distribution, 2017, 11(16): 4096-4103.).

[0006] CN110686899B discloses a method for identifying noise sources in internal combustion engines. This method employs improved variational mode decomposition to adaptively acquire signal components in different frequency bands from the noise signal, and then uses wavelet transform to perform time-frequency analysis on the decomposed noise components. However, this scheme still has limitations in identifying non-order noise sources. When faced with complex non-stationary signals, the use of amplitude spectral entropy as the fitness function may lead to insufficient extraction of non-stationary features.

[0007] Therefore, we consider providing a method that can adaptively and accurately separate complex noise signals from internal combustion engines. Summary of the Invention

[0008] The present invention addresses the shortcomings of the prior art by providing a multi-parameter optimized method for separating noise sources in internal combustion engines. This method employs an improved AO algorithm for adaptive noise decomposition and can combine time-frequency analysis, coherence analysis, and operating condition comparison to accurately separate various types of noise sources, such as combustion noise, piston knocking noise, and gear meshing noise.

[0009] This invention provides a multi-parameter optimized method for separating noise sources in an internal combustion engine, comprising: Acoustic and vibration signals of an internal combustion engine under typical operating conditions and reverse driving conditions are collected; the collected acoustic signals are processed by detrending term processing and bandpass filtering preprocessing to obtain preprocessed noise signals; Using the preprocessed noise signal as the input of variational mode decomposition (VMD), and the number of VMD modes K and the penalty factor α as optimization variables, the improved AO optimization algorithm combined with the energy-weighted fuzzy entropy fitness function is used to adaptively optimize the VMD parameters and perform signal decomposition to obtain several intrinsic mode function (IMF) components. Based on the obtained IMF components, combined with time-frequency analysis, coherence analysis and prior knowledge, non-order noise, periodic impulse noise and order noise are separated and identified to complete the noise source separation. The improvements of the improved AO optimization algorithm over the standard AO algorithm include: By introducing dynamic perturbation factors and adaptive probability coefficients, a balance is struck between global search and local exploitation capabilities. The energy-weighted fuzzy entropy fitness function calculates a weighted average entropy value as a fitness index by combining the energy distribution of IMF components with the fuzzy entropy metric.

[0010] Furthermore, the preprocessing of the acquired acoustic signals, including detrending and bandpass filtering, specifically includes: The original acoustic signal was fitted using the least squares method to eliminate the trend term; and The detrended signal is bandpass filtered to analyze signals in the frequency range of 20-5000Hz.

[0011] Furthermore, the improvements to the improved AO optimization algorithm include: In the full search phase, the pharmacokinetic-based search decay model was modified and a probability coefficient P was introduced. The drug concentration decay formula was also modified to incorporate a sinusoidal perturbation factor. ; in, f This is the current number of calculations. This represents the maximum number of calculations. The original formula for the probability coefficient P is modified to a cosine form: .

[0012] Furthermore, the calculation method of the energy-weighted fuzzy entropy fitness function includes: Reconstructing the state space vector: the first state space vector obtained by using VMD decomposition. i IMF components U IMF Refactoring L-m+1 indivual m dimensional vector X (i) ; Construct the distance matrix and calculate the fuzzy similarity matrix; Define an average similarity function and, in conjunction with each IMF component, calculate the energy-weighted fuzzy entropy: ; in This represents the energy of the i-th component.

[0013] Furthermore, the separation of noise sources, based on the obtained IMF components and combined with time-frequency analysis, coherence analysis, and prior knowledge, to separate and identify non-order noise, periodic impulse noise, and order noise specifically includes: The time-frequency characteristics of the IMF component are obtained by wavelet time-frequency transform. The main frequency bands and time-domain characteristics of the noise component are analyzed. The obtained vibration signal and noise signal are subjected to coherence analysis to identify non-order noise signal. By comparing the spectral density of typical operating conditions and reverse drag operating conditions, and through vibration signal coherence analysis, the periodic characteristics of piston knocking noise and combustion impact noise were identified; and Based on the engine's structural characteristics and the rotational speed ratio of rotating parts, order noise is identified.

[0014] The beneficial effects of this invention are as follows: The proposed noise source separation method simultaneously optimizes the number of modes and the penalty factor through an improved AO (IAO) algorithm, overcoming the shortcomings of existing methods that only optimize a single parameter or ignore the effects of parameter coupling. It employs energy-weighted fuzzy entropy as the fitness function to ensure that the decomposed components accurately correspond to the physical noise sources. Furthermore... By combining time-frequency analysis, coherence analysis, and operating condition comparison, multiple types of noise sources, such as combustion noise, piston knocking noise, and gear meshing noise, can be accurately separated. Attached Figure Description

[0015] Figure 1 This is a flowchart of the multi-parameter optimized internal combustion engine noise source separation method described in Embodiment 1 of the present invention; Figure 2 This is a simulation signal waveform diagram used for verification, separated by the multi-parameter optimized internal combustion engine noise source separation method described in this invention in Example 2; Figure 3 The figure shows the separation results obtained by using the multi-parameter optimized internal combustion engine noise source separation method described in this invention; (a) shows the separated IMF components after using the method proposed in this invention, and (b) shows the fitness function obtained using the method proposed in this invention. The horizontal axis is the number of iteration steps, and the vertical axis is the optimal fitness function value. Figure 4 The figure shows the separation results obtained using the original AO algorithm; (a) shows the separated IMF components after VMD optimization using the original AO algorithm, and (b) shows the fitness function obtained after VMD optimization using the original AO algorithm. The horizontal axis is the number of iteration steps, and the vertical axis is the optimal fitness function value. Figure 5 The figure shows the separation results obtained by using envelope spectral entropy as the fitness function; Figure (a) shows the separated IMF components after VDM optimization using envelope spectral entropy as the fitness function, and Figure (b) shows the fitness function obtained after VDM optimization using envelope spectral entropy as the fitness function. The horizontal axis is the number of iteration steps, and the vertical axis is the optimal fitness function value. Detailed Implementation

[0016] To make the objectives, technical solutions, beneficial effects, and significant advancements of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings provided in the examples of the present invention. Obviously, all 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.

[0017] In the description of this application, unless otherwise expressly specified and limited, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance; the term "multiple" refers to two or more; unless otherwise specified or explained, the terms "connected," "fixed," etc., should be interpreted broadly. For example, "connected" can be a fixed connection, a detachable connection, an integral connection, or an electrical connection; "connected" can be a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0018] Example 1 like Figure 1 As shown, a multi-parameter optimized method for separating noise sources in an internal combustion engine includes: Step 1: Acquisition of acoustic and vibration signals from the internal combustion engine.

[0019] S11: Acoustic signal acquisition: Use the five-point method to set up microphones to acquire the surface-radiated noise signal of the engine.

[0020] S12: Simultaneously acquire vibration signals of key components and cylinder pressure data of the final cylinder of a multi-cylinder engine under typical operating and reverse driving conditions, while simultaneously acquiring acoustic signals, for subsequent verification and analysis. Key components include the cylinder head and block, gear chamber cover, oil pan, crankcase, and other typical sources of significant noise. At the same time, data is acquired for each engine speed under reverse driving conditions to subsequently isolate combustion excitation and obtain pure mechanical noise baseline data. This includes using three-dimensional acceleration sensors in the engine block and cylinder head to acquire vibration signals, and a pressure sensor in the sixth cylinder to measure cylinder pressure. Data is also acquired under reverse driving conditions to isolate combustion excitation and obtain pure mechanical noise baseline data.

[0021] Step Two: Perform detrending and bandpass filtering preprocessing on the raw noise signal acquired in Step One to remove baseline drift and low-frequency interference, providing high-quality data input for subsequent signal decomposition. This includes: S21: The original signal is detrended using the least squares fitting method to eliminate the trend term and reduce signal acquisition error.

[0022] S22: Bandpass filtering is applied to the detrended signal for analysis of the main frequency bands.

[0023] Step 3: Using the preprocessed noise signal from Step 2 as input to Variational Mode Decomposition (VMD), the preset parameters in VMD as optimization variables, and the weighted fuzzy entropy as the fitness function, an improved AO algorithm is used to perform adaptive noise decomposition (AVMD decomposition) to obtain the decomposed signal components. The preset parameters are the number of modes K and the penalty factor α.

[0024] S31: Use the noise signal preprocessed in step two as the input to VMD to construct an improved AO optimization algorithm model.

[0025] Define the optimization variables as the number of modes K and the penalty factor α in VMD, and set their search range.

[0026] An improved AO optimization algorithm model was implemented in Matlab to cyclically call and control the VMD decomposition process.

[0027] S32: Based on the characteristics of the acoustic and vibration signals of internal combustion engines, the search mechanism of the existing standard AO algorithm is improved to form the IAO optimized algorithm. The improvements of the IAO optimized algorithm to the standard AO algorithm are as follows: In the full search phase, the pharmacokinetic-based search decay model was modified and a probability coefficient P was introduced. The standard AO algorithm is based on pharmacokinetic principles and takes into account the decrease in drug concentration over time, as shown in Equation 1. (1) In the formula, C is the drug concentration and k is the rate constant. Solving this differential equation yields Equation 2: (2) Based on this model, assuming an initial drug concentration of 1 and a drug decay rate of 4, we obtain Equation 3: (3) In the formula and These represent the current number of calculations and the maximum number of calculations, respectively.

[0028] To expand the search range, a sinusoidal perturbation factor is introduced into Equation 3, which is modified as shown in Equation 4: (4) Furthermore, considering the varying drug dosages and durations due to differences in patients' conditions and physical states, the algorithm introduces a probability coefficient P to represent this variability, as shown in Equation 5: (5) The probability coefficient P determines whether the population particles should continue with a full search. However, when applied to the optimization of more complex acoustic and vibration signal separation parameters, this formula causes the algorithm to enter the local search stage too early, affecting the accuracy of the final result. Therefore, it is modified to a cosine form, as shown in Equation 6: (6) S33: Construct a fitness function based on weighted fuzzy entropy.

[0029] To overcome the insensitivity of traditional entropy values ​​to non-stationary signals, this embodiment employs energy-weighted fuzzy entropy as the fitness function to guide the algorithm in finding the essential mode that best reflects the signal characteristics. Specifically, this includes calculating the initial local optimum using the weighted fuzzy entropy of the intrinsic mode function (IMF) as the fitness function. best_K and best_a The weighted fuzzy entropy is calculated as follows: 1) Reconstructing the state space vector: using the i-th IMF component obtained from VMD decomposition. U IMF Reconstruct L-m+1 m-dimensional vectors X ( i ): (7) (8) In the formula: u(i) is the i-th sampling point of the IMF component; u0(i) is the mean of the i-th reconstruction vector, which is used to eliminate the DC component of the reconstruction vector.

[0030] 2) Constructing the distance matrix: Based on the reconstructed vectors, construct the distance matrix from... arrive Distance matrix : (9) 3) Calculate the fuzzy similarity matrix: based on fuzzy functions Calculate the distance matrix Similarity matrix : (10) in η and r Representing fuzzy functions respectively Boundary gradient parameters and similarity tolerance of fuzzy membership functions; It is a Gaussian fuzzy membership function.

[0031] 4) Define the average similarity function: (11) 5) Fuzzy entropy is defined as: (12) In the formula, N This represents the number of sampling points in the time series.

[0032] The final weighted fuzzy entropy calculation formula for the noise decomposition components is: (13) In the formula, This represents the energy of the i-th component.

[0033] S34: Perform adaptive decomposition to obtain the decomposed components of the signal.

[0034] The improved AO optimization algorithm model is set to a population size of 50 and a maximum number of iterations of 30. A search strategy is selected based on the probability coefficient P determined by Equation 6, VMD is invoked for decomposition, and the fitness function value is calculated according to Equation 12, continuously updating the local optimum. best_K and best_a And the global optimal solution. Set the convergence condition to reach the preset maximum number of iterations.

[0035] Repeat steps S32-34 until the convergence condition is met, then output the optimal solution. best_K and best_a The obtained optimal solution is then used to perform VMD decomposition on the original signal to obtain several IMF components, thus completing the adaptive separation of the noise signal.

[0036] Step 4: Based on the decomposition results of Step 3, wavelet time-frequency transform is used to obtain the time-frequency characteristics of the components. The main frequency bands and time-domain characteristics of the noise components are analyzed. The vibration signal and noise signal obtained in Step 1 are subjected to coherence analysis to identify non-order noise signals. By combining the spectrum comparison of typical working conditions and reverse driving conditions obtained in Step 1 with the coherence analysis of vibration signals, periodic characteristic signals similar to piston knocking noise and combustion impact noise are identified. Order noise signals are identified through prior knowledge of the engine, structural characteristics, and the rotational speed ratio of rotating parts. Thus, the separation and identification of noise sources are completed.

[0037] Example 2 To verify the effectiveness of the method described in this invention, this embodiment constructs a simulated signal for verification.

[0038] Based on the working principle of internal combustion engines, combustion noise and piston knocking noise have a significant impact on surface radiated noise during operation. Furthermore, these two noises exhibit distinct periodic impact characteristics; therefore, the analog signal must include periodic pulse signals. Additionally, the engine has other sound sources such as rotating gear meshing and component vibration, and the test environment also contains interfering noise. Therefore, several fixed-frequency sinusoidal signals are used to represent component vibration, and a segment of Gaussian noise is used to represent random noise to test the robustness of AVMD decomposition. To detect whether similar frequency band signals will cause mode aliasing during the decomposition process, the pure tone signal can include two sinusoidal signals of similar frequency bands. Based on the above considerations, such as... Figure 2 As shown, this embodiment constructs an analog signal. S To characterize the engine's sound signal, this signal is... S1 (20Hz) S2 (80Hz) S3 Composed of three sinusoidal signals (200Hz), S4 It is a periodic pulse signal located in the 300Hz frequency band. It is a random noise signal.

[0039] Three different algorithms were used to process the above analog signals. S The decomposition process is performed, and the decomposition results are compared and analyzed with the original components. The decomposition results after the method proposed in this invention are as follows: Figure 3 As shown in Figures (a) and (b), the VMD decomposition results after optimization using the original AO algorithm are as follows: Figure 4 Figures (a) and (b) are shown; the decomposition results after VDM optimization using envelope spectral entropy as the fitness function are as follows. Figure 5 As shown in Figures (a) and (b).

[0040] As can be observed from the above figures, the improved AO algorithm used in this invention can improve the separation accuracy and overall convergence speed compared to the original AO algorithm; moreover, it can be confirmed that using energy-weighted fuzzy entropy, compared to the classical fitness function envelope spectrum entropy, can avoid problems such as under-decomposition and can also improve the separation accuracy.

[0041] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style of the specification is merely for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in the embodiments can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A multi-parameter optimized method for separating noise sources in an internal combustion engine, characterized in that, include: Acoustic and vibration signals of internal combustion engines under typical operating conditions and reverse driving conditions were collected. The acquired acoustic signal is subjected to detrending term processing and bandpass filtering preprocessing to obtain the preprocessed noise signal; Using the preprocessed noise signal as the input of variational mode decomposition (VMD), and the number of VMD modes K and the penalty factor α as optimization variables, the improved AO optimization algorithm combined with the energy-weighted fuzzy entropy fitness function is used to adaptively optimize the VMD parameters and perform signal decomposition to obtain several intrinsic mode function (IMF) components. Based on the obtained IMF components, combined with time-frequency analysis, coherence analysis and prior knowledge, non-order noise, periodic impulse noise and order noise are separated and identified to complete the noise source separation. The improvements of the improved AO optimization algorithm over the standard AO algorithm include: By introducing dynamic perturbation factors and adaptive probability coefficients, a balance is struck between global search and local exploitation capabilities. The energy-weighted fuzzy entropy fitness function calculates a weighted average entropy value as a fitness index by combining the energy distribution of IMF components with the fuzzy entropy metric.

2. The multi-parameter optimized method for separating internal combustion engine noise sources according to claim 1, characterized in that, The preprocessing of the acquired acoustic signals, including detrending and bandpass filtering, specifically includes: The original acoustic signal was fitted using the least squares method to eliminate the trend term; and The detrended signal is bandpass filtered to analyze signals in the frequency range of 20-5000Hz.

3. The multi-parameter optimized method for separating internal combustion engine noise sources according to claim 1, characterized in that, The improvements to the AO optimization algorithm include: In the full search phase, the pharmacokinetic-based search decay model was modified and a probability coefficient P was introduced. The drug concentration decay formula was also modified to incorporate a sinusoidal perturbation factor. ； in, f This is the current number of calculations. This represents the maximum number of calculations. The original formula for the probability coefficient P is modified to a cosine form: 。 4. The multi-parameter optimized method for separating internal combustion engine noise sources according to claim 1, characterized in that, The calculation method of the energy-weighted fuzzy entropy fitness function includes: Reconstructing the state space vector: using the i-th IMF component obtained from VMD decomposition. U IMF Reconstruct L-m+1 m-dimensional vectors X ( i ); Construct the distance matrix and calculate the fuzzy similarity matrix; Define an average similarity function and, in conjunction with each IMF component, calculate the energy-weighted fuzzy entropy: ； in This represents the energy of the i-th component.

5. The multi-parameter optimized method for separating internal combustion engine noise sources according to claim 1, characterized in that, The process of separating and identifying non-order noise, periodic impulse noise, and order noise based on the obtained IMF components, combined with time-frequency analysis, coherence analysis, and prior knowledge, to complete noise source separation specifically includes: The time-frequency characteristics of the IMF component are obtained by wavelet time-frequency transform. The main frequency bands and time-domain characteristics of the noise component are analyzed. The obtained vibration signal and noise signal are subjected to coherence analysis to identify non-order noise signal. By comparing the spectral density of typical operating conditions and reverse drag operating conditions, and through vibration signal coherence analysis, the periodic characteristics of piston knocking noise and combustion impact noise were identified; and Based on the engine's structural characteristics and the rotational speed ratio of rotating parts, order noise is identified.

Images