A noise control method, system, device and storage medium of a micro engine

Abstract

The application discloses a noise control method, system and device of a micro engine and a storage medium, belongs to the technical field of engine noise control, and solves the technical problem that the micro engine is difficult to achieve an ideal noise reduction effect. The method is: collecting intake / exhaust noise, mechanical noise and combustion noise data in the working process of the micro engine, and performing empirical mode decomposition to decompose the noise signal into a plurality of IMF components of intrinsic mode functions; an LSTM neural network model is constructed; the LSTM neural network model is trained to obtain an EMD-EGMGWO-LSTM model; in the working process of the micro engine, noise data is collected in real time and input into the trained EMD-EGMGWO-LSTM model to obtain a predicted value of the noise data; and an anti-phase noise control signal opposite in phase and matched in amplitude is generated in real time to achieve an efficient active noise reduction effect. The control strategy can be continuously optimized and adjusted, and the noise reduction effect is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engine noise control, more particularly, it relates to a micro engine noise control method, system, device and storage medium. BACKGROUND

[0002] Micro engines have a wide range of applications in many fields due to their small size and high power density. However, the compact design makes noise control a major problem.

[0003] Traditional physical noise reduction methods, such as soundproofing and mufflers, are difficult to implement effectively without affecting performance due to space constraints and weight considerations. In addition, the dynamic changes in the working state of micro engines make the noise characteristics complex and variable. Although the traditional active noise control (ANC) strategy based on FxLMS algorithm relies on experience and experimental data, it is limited by the complexity of secondary path modeling, high algorithm complexity, and insufficient adaptability to changes in engine dynamic noise characteristics, making it difficult to achieve ideal noise reduction effect. SUMMARY

[0004] The technical problem to be solved by the present application is to overcome the above-mentioned deficiencies of the prior art. The first object of the present application is to provide a micro engine noise control method.

[0005] The second object of the present application is to provide a micro engine noise control system.

[0006] The third object of the present application is to provide a computer device.

[0007] The fourth object of the present application is to provide a computer storage medium.

[0008] In order to achieve the above-mentioned first object, the present application provides a micro engine noise control method, comprising the following steps:

[0009] Step 1. Data acquisition and preprocessing, real-time acquisition of intake / exhaust noise, mechanical noise and combustion noise data during the working process of the micro engine through the noise sensor, and empirical mode decomposition, decomposition of the noise signal into several IMF components of the intrinsic mode function, to obtain different frequency components of the micro engine noise;

[0010] Step 2. Constructing an LSTM neural network model according to the characteristics of the micro engine noise, the LSTM neural network model including an input layer, a hidden layer and an output layer; the input layer is responsible for receiving the IMF components of the intrinsic mode function or the time series segments of the original noise data after empirical mode decomposition preprocessing, the hidden layer extracts noise features through nonlinear transformation, and the output layer outputs noise prediction values or control signals;

[0011] Step 3. Training and optimization, based on the noise data of the micro engine under multiple working conditions, the LSTM neural network model is trained, and the number of LSTM hidden layer neurons and learning rate are optimized through EGMGWO, obtaining the EMD-EGMGWO-LSTM model capable of predicting the noise of the micro engine, at the same time, L2 regularization is adopted to prevent EMD-EGMGWO-LSTM model overfitting, and the effective generalization of EMD-EGMGWO-LSTM model is realized;

[0012] Step 4. Real-time noise control, in the working process of the micro engine, real-time noise data is collected and input into the trained EMD-EGMGWO-LSTM model to obtain the predicted value of the noise data; based on the predicted value, an anti-phase noise control signal with opposite phase and matched amplitude is generated in real time, and through the vibration suppression device, an anti-phase sound wave or vibration that cancels the original noise waveform is generated, thereby realizing efficient active noise reduction effect.

[0013] As a further improvement, the intake / exhaust noise includes: exhaust pipe periodic noise, resonance noise in the engine pipe;

[0014] The base frequency expression of the engine exhaust pipe periodic noise can be expressed as:

[0015]

[0016] In formula (1), P f represents the periodic frequency of the noise; N m represents the number of working cylinders in the engine; k n represents the maximum speed of the engine under ideal conditions; λ k represents the engine cylinder displacement;

[0017] The resonance noise in the engine pipe can be expressed as:

[0018]

[0019] In formula (2), H f represents the resonance noise of the engine intake / exhaust; S p represents the cross-sectional area of the intake / exhaust pipe; V p represents the total volume of all cylinders in the engine; L p represents the length of the intake / exhaust pipe.

[0020] Further, the engine combustion noise is the noise caused by pressure fluctuation and vibration due to fuel combustion in the cylinder when the engine is in working condition, which is described as follows:

[0021] The frequency spectrum density C(f) of the combustion noise can be estimated by formula (3):

[0022] C(f) = |H(f) 2 ·P(f) (3)

[0023] Where H(f) is the transfer function of the combustion noise, which is related to the engine structure and the shape of the combustion chamber; P(f) is the spectral density of the combustion pressure;

[0024] Combustion noise sound pressure level L rz It can be described by equation (4):

[0025]

[0026] Where p rz (t) is the time-domain signal of the combustion noise, and T is the measurement time.

[0027] Further, the mechanical noise mainly comes from the friction, vibration and impact of the moving parts inside the engine, and the mechanical noise includes structural dynamic noise and gear noise.

[0028] The structural dynamic noise refers to the noise excited by the friction and impact of moving parts and the noise produced by the inherent vibration of parts excited by mechanical forces, and its sound pressure level can be estimated by equation (5):

[0029]

[0030] Where p i (t) is the time-domain signal of the i-th noise source, and n is the number of noise sources.

[0031] Gear noise sound pressure level L clz It can be estimated by equation (6):

[0032]

[0033] In the formula, p cl (t) is the time-domain signal of the gear noise.

[0034] Further, the process of empirical mode decomposition includes:

[0035] The recursive average value is calculated by equation (7):

[0036]

[0037] In equation (7), IMF represents the recursive average value of the interpolation function extreme value; f max and f min represent the envelope lines of the maximum and minimum values of the function, respectively.

[0038] Through the average value calculation, the residual amount of each noise point compared to the extreme point can be obtained:

[0039] r m = x(t) - IMF (8)

[0040] In formula (8): r m represents the residual amount of each noise point relative to the current IMF, that is, the difference between the original signal and the IMF; x(t) represents the original noise signal; the formula is used to update the signal continuously to extract all IMF components;

[0041] The characteristics represented by the modal classification can be obtained by decomposing and screening at different frequencies, and the screening mechanism is:

[0042]

[0043] In formula (9): μ sd represents the residual ratio of the screening mechanism in any discrete sequence, which is used to determine whether the extraction of the IMF component is completed; h k-1 (t) and h k (t) represent the screening components of two adjacent regions, that is, the intermediate results in the IMF extraction process, respectively;

[0044] Under the screening mechanism, the output value of the engine noise signal can be obtained, and at this time the engine noise signal under different decomposition units can be represented as:

[0045]

[0046] In formula (10), x(t) represents the original noise signal; IMF i represents the i-th IMF component, which represents the modal characteristics of the signal at different frequencies; r n represents the last residual component, that is, the signal part remaining after all IMF components are extracted.

[0047] Further, the principle of realizing the optimization of the number of LSTM hidden layer neurons and the learning rate in step 3 through EGMGWO is that the number of LSTM hidden layer neurons and the learning rate are optimized through the GWO algorithm, and the ability of the GWO algorithm to obtain optimal parameters is improved by introducing a dynamic adjustment convergence factor and a Gaussian variation local search mechanism, the process is as follows:

[0048] Step 31. Algorithm initialization, including:

[0049] 1) Population initialization, set population size N, search space dimension d, maximum iteration number T max ; randomly generate initial population position X i = (x i1 , x i2 , …, x id ), where i = 1, 2, …, N;

[0050] 2) Parameter initialization, initialize convergence factor a init , decay constant T delay , dynamic adjustment parameter λ min , dynamic adjustment parameter λ max , dynamic adjustment parameter β; initialize Gaussian variation parameters, mean μ = 0, standard deviation σ; initialize the positions of Alpha, Beta and Delta of three gray wolves and their fitness values; set the acceptance probability p of Gaussian variation local search accept ;

[0051] Step 32. Fitness evaluation, for each individual X i , use the LSTM hyperparameter configuration it represents to train the LSTM neural network model, and combine the engine noise data processed by empirical mode decomposition to make prediction; calculate the mean square error (MSE) between the prediction result and the real noise data as the fitness value f(X i ):

[0052]

[0053] Where y k is the real noise data, is the prediction result of the LSTM neural network model, and n is the number of data points;

[0054] Step 33. Position update, including:

[0055] 1) Calculate the coefficient vector and For each individual X i , generate random numbers r1, r2 ∈ [0, 1]; calculate the nonlinearly decaying convergence factor:

[0056]

[0057] Calculate the coefficient vector:

[0058] A i = 2a(t)·r i -a(t), C i = 2·sin(2πr2) (13);

[0059] 2) Calculate the dynamic adjustment factor, calculate the dynamic adjustment factor according to the fitness of the population and the number of iterations:

[0060]

[0061] 3) Update the position of gray wolf individual, for each individual X i , according to its current position X i (t), the position of Alpha Xα , the position X of Beta β , the position X of Delta δ , and the coefficient vector A i , C i , and the dynamic adjustment factor λ, update the individual position:

[0062]

[0063] 4) Gaussian mutation local search, for each dimension x i (t+1) of the updated individual X ij (t+1), apply Gaussian mutation:

[0064]

[0065] where N(0, 1) represents the standard normal distribution;

[0066] Evaluate the fitness of the newly generated candidate solution , and obtain

[0067] Update the individual position according to the set acceptance probability p accept :

[0068]

[0069] where r is a randomly generated number, r ∈ [0, 1];

[0070] Step 34. Boundary processing, set the value range of the hyperparameters of the LSTM neural network model as [L j , U j ], where L j and U j are the lower and upper bounds of the jth hyperparameter; after each position update, boundary checking and processing need to be performed on the updated individual position X i (t+1) = (x i1 (t+1), x i2 (t+1), …, x id (t+1));

[0071] For each dimension j = 1, 2, …, d, perform the following operations:

[0072]

[0073] The individual position X i (t+1) after boundary processing can satisfy the value range constraints of all hyperparameters;

[0074] Step 35. Iteration termination condition, repeat the fitness evaluation, position update and boundary processing until the maximum iteration number T is reached max ;

[0075] Step 36. Optimal solution output, according to the fitness value, output the optimal individual position X best and its fitness value f(X best ), using X best represent the hyperparameter configuration of the LSTM neural network model to train the final EMD-EGMGWO-LSTM model.

[0076] Further, the LSTM neural network model is constructed and grey wolf optimization as follows:

[0077] A micro engine noise suppression model is built, and the neural network has three layers, one input layer, one hidden layer and one output layer;

[0078] The input layer has 12 nodes, and a linear activation function is selected;

[0079] The neural network structure used in the hidden layer is a standard LSTM layer, without additional activation function, and the number of hidden layer nodes is determined according to the GWO algorithm optimization;

[0080] The output layer has 1 node, the Epoch is set to 500, and the gradient descent optimization algorithm selects Adam;

[0081] The initial parameters of the GWO algorithm are: the number of wolf populations is 50, the maximum iteration number is 500, the dimension is 3, the learning rate search range is 0.001-0.1, the hidden layer node search range is 10-100, and the L2 regularization coefficient search range is 0.001-0.1.

[0082] In order to achieve the above-mentioned second purpose, the present application provides a micro engine noise control system, comprising:

[0083] The acquisition module is used for data acquisition and preprocessing, and real-time acquisition of intake / exhaust noise, mechanical noise and combustion noise data in the working process of the micro engine through a noise sensor, and empirical mode decomposition is performed to decompose the noise signal into a plurality of intrinsic mode function IMF components to obtain different frequency band components of the micro engine noise.

[0084] The model construction module is used for constructing an LSTM neural network model according to the characteristics of the micro engine noise, wherein the LSTM neural network model comprises an input layer, a hidden layer and an output layer; the input layer is responsible for receiving the intrinsic mode function IMF component or the time series segment of the original noise data after empirical mode decomposition preprocessing, the hidden layer extracts noise features through nonlinear transformation, and the output layer outputs noise prediction values or control signals;

[0085] The training module is used for training and optimization, LSTM neural network model is trained based on micro engine noise data in multiple working conditions, the number of LSTM hidden layer neurons and the learning rate are optimized through EGMGWO, the EMD-EGMGWO-LSTM model capable of predicting micro engine noise is obtained, meanwhile, L2 regularization is adopted to prevent EMD-EGMGWO-LSTM model from overfitting, and effective generalization of the EMD-EGMGWO-LSTM model is realized.

[0086] The control module is used for realizing real-time noise control, noise data is collected in real time during the working process of the micro engine and input into the trained EMD-EGMGWO-LSTM model, the predicted value of the noise data is obtained; based on the predicted value, the anti-phase noise control signal with opposite phase and matched amplitude is generated in real time, and the reverse sound wave or vibration that cancels the original noise waveform is generated through the vibration suppression device, so that the efficient active noise reduction effect is realized.

[0087] In order to realize the above-mentioned purpose three, the application provides a computer device, comprising a memory and a processor, the memory stores a computer program, and the processor realizes the above-mentioned micro engine noise control method when executing the computer program.

[0088] In order to realize the above-mentioned purpose four, the application provides a computer readable storage medium, which stores a computer program, and the computer program realizes the above-mentioned micro engine noise control method when executed by a processor.

[0089] Advantages

[0090] Compared with the prior art, the application has the advantages that:

[0091] The application can realize intelligent analysis and control of noise by collecting and processing noise signals of the micro engine in real time, can improve the prediction and suppression ability of nonlinear noise through adaptive learning and optimization of the neural network model, can adapt to the noise control demand under various complex working conditions, and can continuously optimize and adjust the control strategy through the real-time feedback mechanism to improve the noise reduction effect. BRIEF DESCRIPTION OF DRAWINGS

[0092] Fig. 1 The flowchart of the application is shown in the figure.

[0093] Fig. 2 The flowchart of the improved Gaussian variation grey wolf optimization algorithm in the application is shown in the figure. DETAILED DESCRIPTION

[0094] The application will be further described below in combination with specific embodiments in the figures.

[0095] Referring to Figs. 1-2 A noise control method of a micro engine, comprising the following steps 1-4:

[0096] Step 1. Data acquisition and preprocessing, real-time acquisition of intake / exhaust noise, mechanical noise and combustion noise data during the working process of the micro engine by a noise sensor such as a sound level meter, and empirical mode decomposition EMD, decomposing the noise signal into several intrinsic mode function IMF components to obtain different frequency components of the micro engine noise.

[0097] Step 2. Constructing an LSTM neural network model according to the characteristics of the micro engine noise, which includes an input layer, a hidden layer and an output layer; the input layer is responsible for receiving the intrinsic mode function IMF component or the time series segment of the original noise data after empirical mode decomposition EMD preprocessing, the hidden layer extracts noise features through nonlinear transformation, and the output layer outputs noise prediction values or control signals.

[0098] Step 3. Training and optimization, training the LSTM neural network model based on the micro engine noise data under multiple working conditions, optimizing the number of LSTM hidden layer neurons and learning rate through EGMGWO, and obtaining an EMD-EGMGWO-LSTM model capable of predicting the micro engine noise. At the same time, L2 regularization is used to prevent EMD-EGMGWO-LSTM model overfitting, and effective generalization of EMD-EGMGWO-LSTM model is realized.

[0099] Step 4. Real-time noise control, real-time acquisition of noise data during the working process of the micro engine and input into the trained EMD-EGMGWO-LSTM model to obtain the prediction value of the noise data; based on the prediction value, an anti-phase noise control signal with opposite phase and matched amplitude is generated, and through a vibration suppression device, an anti-phase sound wave or vibration is generated to offset the original noise waveform, thereby realizing efficient active noise reduction effect.

[0100] The noise of the engine mainly comes from the impact of gas on the valve during the intake / exhaust process, mechanical noise and combustion noise.

[0101] Among them, the intake / exhaust noise includes: exhaust pipe periodic noise, resonance noise in the engine pipe. The engine intake / exhaust noise can be described by equations (1)-(2).

[0102] The expression of the fundamental frequency of the engine exhaust pipe periodic noise can be expressed as:

[0103]

[0104] In equation (1), Pf denotes the periodic frequency of the noise; N m denotes the number of working cylinders in the engine; k n denotes the maximum speed of the engine in ideal state; λ k denotes the engine cylinder displacement.

[0105] The resonance noise in the engine pipe can be expressed as:

[0106]

[0107] In formula (2), H f denotes the resonance noise of the engine intake / exhaust; S p denotes the cross-sectional area of the intake / exhaust pipe; V p denotes the total volume of all cylinders in the engine; L p denotes the length of the intake / exhaust pipe.

[0108] The engine combustion noise is the noise caused by the pressure fluctuation and vibration due to the combustion of fuel in the cylinder when the engine is in working state, which can be described by formulas (3)-(4):

[0109] The frequency spectrum density C(f) of the combustion noise can be estimated by formula (3):

[0110] C(f) = |H(f) 2 ·P(f) (3)

[0111] wherein H(f) is the transfer function of the combustion noise, which is related to the factors of engine structure and combustion chamber shape; P(f) is the frequency spectrum density of the combustion pressure.

[0112] The sound pressure level L rz of the combustion noise can be described by formula (4):

[0113]

[0114] wherein p rz (t) is the time domain signal of the combustion noise, and T is the measurement time.

[0115] The mechanical noise mainly comes from the friction, vibration and impact of the moving parts inside the engine, and the mechanical noise includes structural dynamic noise and gear noise, which can be described by formulas (5)-(6):

[0116] The structural dynamic noise refers to the noise excited by the friction and impact of moving parts and the noise produced by the inherent vibration of parts excited by mechanical forces, and its sound pressure level can be estimated by formula (5):

[0117]

[0118] where p i (t) is the time-domain signal of the ith noise source, and n is the number of noise sources.

[0119] The sound pressure level L clz It can be estimated by formula (6):

[0120]

[0121] where p cl (t) is the time-domain signal of the gear noise.

[0122] The process of empirical mode decomposition (EMD) can be described using formulas (7)-(10):

[0123] The recursive mean value is calculated by formula (7):

[0124]

[0125] In formula (7), IMF represents the recursive mean value of the interpolation function extreme value; f max and f min represent the envelope lines of the function's maximum and minimum values, respectively.

[0126] The residual amount of each noise point relative to the extreme point can be obtained by mean value calculation:

[0127] r m = x(t) - IMF (8)

[0128] In formula (8), r m represents the residual amount of each noise point relative to the current IMF, i.e., the difference between the original signal and the IMF; x(t) represents the original noise signal; this formula is used to continuously update the signal to extract all IMF components.

[0129] By decomposing and screening the characteristics represented by the modal classification at different frequencies, the screening mechanism can be obtained as:

[0130]

[0131] In formula (9), μ sd represents the residual amount ratio of the screening mechanism within any discrete sequence, which is used to determine whether the extraction of the IMF component is complete; h k-1 (t) and h k (t) represent the screening components of two adjacent regions, i.e., the intermediate results in the IMF extraction process.

[0132] Under the screening mechanism, the output value of the engine noise signal can be obtained, and at this time, the engine noise signal under different decomposition units can be represented as:

[0133]

[0134] In formula (10), x(t) represents the original noise signal; IMF i represents the i-th IMF component, representing the modal characteristics of the signal at different frequencies; r n represents the last residual component, i.e., the signal part remaining after all IMF components are extracted. After EMD processing, the components of the engine noise signal under different empirical modes can be obtained.

[0135] The principle of optimizing the number of LSTM hidden layer neurons and the learning rate in step 3 through EGMGWO is to optimize the number of LSTM hidden layer neurons and the learning rate through GWO algorithm, and to improve the ability of GWO algorithm to obtain optimal parameters by introducing dynamic adjustment of convergence factor and Gaussian variation local search mechanism, to solve the problem of local optimum of traditional GWO algorithm. As shown in Fig. 2 , the process is as follows:

[0136] Step 31. Algorithm initialization, including:

[0137] 1) Population initialization, set population size N, search space dimension d, maximum iteration number T max ; randomly generate initial population position X i =(x i1 ,x i2 ,…,x id ), where i=1,2,…,N.

[0138] 2) Parameter initialization, initialize convergence factor a init , attenuation constant T delay , dynamic adjustment parameter λ min , dynamic adjustment parameter λ max , dynamic adjustment parameter β; initialize Gaussian variation parameters, mean μ=0, standard deviation σ; initialize the positions of Alpha, Beta, Delta of three grey wolves and their fitness values; set the acceptance probability p accept of Gaussian variation local search.

[0139] Step 32. Fitness evaluation, for each individual X i , use the LSTM hyperparameter configuration represented by it to train the LSTM neural network model, and combine the engine noise data processed by empirical mode decomposition EMD for prediction. Calculate the mean square error (MSE) between the prediction result and the true noise data as the fitness value f(X i ):

[0140]

[0141] where y kis real noise data, is the prediction result of the LSTM neural network model, and n is the number of data points.

[0142] Step 33. Position update, including:

[0143] 1) Calculate the coefficient vector and For each individual X i , generate random numbers r1, r2 ∈ [0, 1]; calculate the convergence factor of nonlinear attenuation:

[0144]

[0145] Calculate the coefficient vector:

[0146] A i = 2a(t)·r i -a(t), C i = 2·sin(2πr2) (13);

[0147] 2) Calculate the dynamic adjustment factor, calculate the dynamic adjustment factor according to the population fitness and the number of iterations:

[0148]

[0149] 3) Grey wolf individual position update, for each individual X i , according to its current position X i (t), the position of Alpha X α , the position of Beta X β , the position of Delta X δ , and the coefficient vector A i , C i and the dynamic adjustment factor λ, update the individual position.

[0150]

[0151] 4) Gaussian mutation local search, for each dimension x i (t+1) of the updated individual X ij (t+1), apply Gaussian mutation:

[0152]

[0153] Where N(0, 1) represents the standard normal distribution.

[0154] Evaluate the fitness of the newly generated candidate solution , get

[0155] According to the set acceptance probability paccept Update individual position:

[0156]

[0157] where r is a randomly generated number, r∈[0,1].

[0158] Step 34. Boundary processing, set the value range of the hyperparameters of the LSTM neural network model as [L j , U j ], where L j and U j are the lower and upper bounds of the jth hyperparameter, respectively; after each position update, the updated individual position X i (t+1) = (x i1 (t+1), x i2 (t+1),…, x id (t+1)) needs to be checked and processed.

[0159] For each dimension j = 1, 2, …, d, perform the following operations:

[0160]

[0161] The individual position X i (t+1) after boundary processing can satisfy the value range constraints of all hyperparameters.

[0162] Step 35. Iteration termination condition, repeat the fitness evaluation, position update, and boundary processing until the maximum number of iterations T max is reached.

[0163] Step 36. Optimal solution output, according to the fitness value, output the optimal individual position X best and its fitness value f(X best ), and use X best to represent the hyperparameter configuration of the LSTM neural network model to train the final EMD-EGMGWO-LSTM model.

[0164] LSTM neural network model construction and grey wolf optimization:

[0165] Build a micro engine noise suppression model, set the neural network to have three layers, one input layer, one hidden layer, and one output layer;

[0166] The input layer has 12 nodes and uses a linear activation function;

[0167] The neural network structure used in the hidden layer is a standard LSTM layer without additional activation functions, and the number of hidden layer nodes is determined by optimization according to the GWO algorithm;

[0168] The output layer has one node, the Epoch is set to 500, and the gradient descent optimization algorithm is Adam;

[0169] The initial parameters of the GWO algorithm are: the wolf population number is 50, the maximum iteration number is 500, the dimension is 3, the learning rate search range is 0.001-0.1, the hidden layer node search range is 10-100, and the L2 regularization coefficient search range is 0.001-0.1.

[0170] A noise control system of a micro engine, comprising:

[0171] The acquisition module is used for data acquisition and preprocessing, and real-time acquisition of intake / exhaust noise, mechanical noise and combustion noise data during the working process of the micro engine through a noise sensor, and experience mode decomposition EMD is performed to decompose the noise signal into a plurality of intrinsic mode function IMF components to obtain different frequency band components of the micro engine noise.

[0172] The model construction module is used for constructing an LSTM neural network model according to the characteristics of the micro engine noise, and the LSTM neural network model comprises an input layer, a hidden layer and an output layer; the input layer is responsible for receiving the intrinsic mode function IMF component or the time sequence segment of the original noise data after experience mode decomposition EMD preprocessing, the hidden layer extracts noise features through nonlinear transformation, and the output layer outputs noise prediction values or control signals;

[0173] The training module is used for training and optimization, and the LSTM neural network model is trained based on the micro engine noise data under multiple working conditions, the number of LSTM hidden layer neurons and the learning rate are optimized through EGMGWO, and an EMD-EGMGWO-LSTM model capable of predicting the micro engine noise is obtained, meanwhile, L2 regularization is adopted to prevent the EMD-EGMGWO-LSTM model from overfitting, and effective generalization of the EMD-EGMGWO-LSTM model is realized.

[0174] The control module is used for realizing real-time noise control, and in the working process of the micro engine, noise data is collected in real time and input into the trained EMD-EGMGWO-LSTM model to obtain the prediction value of the noise data; based on the prediction value, an anti-phase noise control signal with opposite phase and matched amplitude is generated in real time, and a reverse sound wave or vibration that cancels out the original noise waveform is generated through a vibration suppression device, thereby realizing efficient active noise reduction effect.

[0175] A computer device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned noise control method of a micro engine when executing the computer program.

[0176] A computer readable storage medium, having stored thereon a computer program, the computer program being executed by a processor to implement the micro-engine noise control method.

[0177] The above merely describes the preferred embodiments of the present application, and it should be pointed out that those skilled in the art can make several modifications and improvements without departing from the structure of the present application, and these will not affect the effect of the present application and the practicality of the patent.

Claims

1. A noise control method for a micro-engine, characterized in that, Includes the following steps: Step 1. Data acquisition and preprocessing: The intake / exhaust noise, mechanical noise and combustion noise data of the micro engine are collected in real time through noise sensors, and empirical mode decomposition (EMD) is performed to decompose the noise signal into several intrinsic mode functions (IMF) components to obtain the different frequency band components of the micro engine noise. Step 2. Construct an LSTM neural network model based on the characteristics of the micro engine noise. The LSTM neural network model includes an input layer, a hidden layer, and an output layer. The input layer is responsible for receiving the IMF components of the intrinsic mode functions after Empirical Mode Decomposition (EMD) preprocessing or time series segments of the original noise data. The hidden layer extracts noise features through nonlinear transformation. The output layer outputs noise prediction values ​​or control signals. Step 3. Training and optimization: The LSTM neural network model is trained based on the noise data of micro engines under multiple working conditions. The number of neurons in the hidden layer of the LSTM is optimized and the learning rate is optimized through EGMGWO to obtain the EMD-EGMGWO-LSTM model that can predict the noise of micro engines. At the same time, L2 regularization is used to prevent the EMD-EGMGWO-LSTM model from overfitting and to achieve effective generalization of the EMD-EGMGWO-LSTM model. Step 4. Real-time noise control: During the operation of the micro engine, noise data is collected in real time and input into the trained EMD-EGMGWO-LSTM model to obtain the predicted value of the noise data. Based on the predicted value, an anti-phase noise control signal with opposite phase and matching amplitude is generated in real time. The anti-sound wave or vibration that cancels out the original noise waveform is generated by the vibration suppression device, thereby achieving efficient active noise reduction. The principle behind optimizing the number of neurons and learning rate in the LSTM hidden layer using GWO in step 3 is as follows: The GWO algorithm is used to optimize the number of neurons and learning rate in the LSTM hidden layer. Furthermore, by introducing a dynamically adjusted convergence factor and a Gaussian mutation local search mechanism, the ability of the GWO algorithm to obtain optimal parameters is improved. The process is as follows: Step 31. Algorithm initialization, including: 1) Population initialization, setting the population size N Search space dimension d Maximum number of iterations Randomly generate initial population locations ,in ; 2) Parameter initialization, initialize convergence factor Attenuation constant Dynamically adjust parameters Dynamically adjust parameters Dynamically adjust parameters Initialize Gaussian mutation parameters, mean Standard deviation Initialize the positions and fitness values ​​of Alpha, Beta, and Delta for three individual gray wolves; set the acceptance probability of Gaussian mutation local search. ; Step 32. Fitness assessment for each individual. An LSTM neural network model is trained using its representative LSTM hyperparameter configuration, and prediction is performed using engine noise data processed by Empirical Mode Decomposition (EMD). The mean square error (MSE) between the prediction result and the actual noise data is calculated as the fitness value. : （11） in, It is real noise data. It is the prediction result of the LSTM neural network model. n It refers to the number of data points; Step 33. Location update, including: 1) Calculate the coefficient vector and For each individual Generate random numbers ; Calculate the convergence factor of the nonlinear decay: （12） Calculate the coefficient vector: ， （13）； 2) Calculate the dynamic adjustment factor based on the population fitness and the number of iterations: （14） 3) Gray wolf individual position update, for each individual According to its current location Alpha's position Beta position The position of Delta and coefficient vector , and dynamic adjustment factor Update individual location: （15） 4) Gaussian mutation local search, for the updated individual Each dimension Apply Gaussian mutation: （16） in, Represents a standard normal distribution; For the newly generated candidate solutions Fitness evaluation was conducted to obtain ; Based on the set acceptance probability Update individual location: （17） in, r The number is randomly generated. ; Step 34. Boundary handling: Assume the range of values ​​for the hyperparameters of the LSTM neural network model is... ,in and They are the first j The lower and upper bounds of each hyperparameter need to be determined; after each position update, the updated individual position needs to be... Perform boundary checks and processing; For each dimension Perform the following operations: （18） Realize the individual position after boundary processing Satisfy the range constraints of all hyperparameter values; Step 35. Set the iteration termination condition, and repeat the fitness evaluation, position update, and boundary handling until the maximum number of iterations is reached. ; Step 36. Optimal Solution Output: Based on the fitness value, output the optimal individual position. and its fitness value ,use The hyperparameter configuration of the representative LSTM neural network model is used to train the final EMD-EGMGWO-LSTM model.

2. The noise control method for a micro-engine according to claim 1, characterized in that, Engine combustion noise is the noise caused by pressure fluctuations and vibrations resulting from fuel combustion within the cylinders when the engine is running. It is described as follows: Spectral density of combustion noise It can be estimated using equation (3): （3） in, It is the transfer function of combustion noise, which is related to factors such as engine structure and combustion chamber shape; It is the spectral density of combustion pressure; Combustion noise sound pressure level It can be described using equation (4): （4） in, It is the time-domain signal of combustion noise. T For measuring time.

3. The noise control method for a micro-engine according to claim 1, characterized in that, Mechanical noise mainly originates from the friction, vibration, and impact of moving parts inside the engine. Mechanical noise includes structural dynamic noise and gear noise. Structural dynamic noise refers to the noise generated by friction and impact of moving parts and the noise generated by the inherent vibration of parts due to mechanical forces. Its sound pressure level is estimated by equation (5): （5） in, It is the first i The time-domain signal of a noise source, n Number of noise sources; sound pressure level of gear noise Estimation is performed using equation (6): （6） In the formula, This is the time-domain signal of gear noise.

4. The noise control method for a micro-engine according to claim 1, characterized in that, The Empirical Mode Decomposition (EMD) process includes: The recursive average is calculated using equation (7): （7） In equation (7): This represents the recursive average of the extrema of the interpolation function; and These represent the envelopes of the function's maxima and minima, respectively. The residual value of each noise point compared to the extreme point can be obtained by averaging: （8） In equation (8): This indicates that each noise point is relative to the current... IMF The residual amount, i.e., the original signal and IMF difference; This represents the original noise signal; this formula is used to continuously update the signal to extract all... IMF Quantity; By decomposing and filtering the characteristics represented by modal classifications at different frequencies, the filtering mechanism is obtained as follows: （9） In equation (9): This represents the residual ratio of the screening mechanism within any discrete sequence, used to determine... IMF Has the extraction of the components been completed? and These represent the filtering components of two adjacent regions, i.e. IMF Intermediate results during the extraction process; Under the filtering mechanism, the output value of the engine noise signal is obtained. At this time, the engine noise signal is represented by different decomposition units as follows: （10） In equation (10), Represents the original noise signal; Indicates the first i indivual IMF The components represent the modal characteristics of a signal at different frequencies; This represents the final residual component, i.e., all IMF The remaining signal portion after component extraction.

5. The noise control method for a micro-engine according to claim 1, characterized in that, The LSTM neural network model construction and gray wolf optimization are as follows: A noise suppression model for a micro engine was built, and the neural network was set to have three layers: an input layer, a hidden layer, and an output layer. The input layer has 12 nodes and uses a linear activation function. The hidden layer uses a standard LSTM layer neural network structure without adding any additional activation functions. The number of nodes in the hidden layer is determined by optimization based on the GWO algorithm. The output layer has one node, the Epoch is set to 500, and the gradient descent optimization algorithm is Adam. The initial parameters of the GWO algorithm are: 50 wolf packs, 500 maximum iterations, 3 dimensions, a learning rate search range of 0.001 to 0.1, a hidden layer node search range of 10 to 100, and an L2 regularization coefficient search range of 0.001 to 0.

1.

6. A noise control system for a micro-engine, characterized in that, include: The acquisition module is used for data acquisition and preprocessing. It collects the intake / exhaust noise, mechanical noise and combustion noise data of the micro engine in real time through noise sensors, and performs empirical mode decomposition (EMD) to decompose the noise signal into several intrinsic mode functions (IMF) components to obtain the different frequency band components of the micro engine noise. The model building module is used to construct an LSTM neural network model based on the characteristics of micro-engine noise. The LSTM neural network model includes an input layer, a hidden layer, and an output layer. The input layer is responsible for receiving the IMF components of the intrinsic mode functions after Empirical Mode Decomposition (EMD) preprocessing or time series segments of the original noise data. The hidden layer extracts noise features through nonlinear transformation, and the output layer outputs noise prediction values ​​or control signals. The training module is used for training and optimization. It trains the LSTM neural network model based on the noise data of micro engines under multiple working conditions. The number of neurons in the hidden layer of the LSTM is optimized and the learning rate is optimized through EGMGWO, so as to obtain the EMD-EGMGWO-LSTM model that can predict the noise of micro engines. At the same time, L2 regularization is used to prevent the EMD-EGMGWO-LSTM model from overfitting and to achieve effective generalization of the EMD-EGMGWO-LSTM model. The control module is used to realize real-time noise control. During the operation of the micro engine, noise data is collected in real time and input into the pre-trained EMD-EGMGWO-LSTM model to obtain the predicted value of the noise data. Based on the predicted value, an anti-phase noise control signal with opposite phase and matching amplitude is generated in real time. The anti-sound wave or vibration that cancels the original noise waveform is generated by the vibration suppression device, thereby achieving efficient active noise reduction effect. The principle behind optimizing the number of neurons and learning rate in the LSTM hidden layer using EGMGW is as follows: The GWO algorithm is used to optimize the number of neurons and learning rate in the LSTM hidden layer. Furthermore, by introducing a dynamically adjusted convergence factor and a Gaussian mutation local search mechanism, the ability of the GWO algorithm to obtain optimal parameters is improved. The process is as follows: Algorithm initialization, including: 1) Population initialization, setting the population size N Search space dimension d Maximum number of iterations Randomly generate initial population locations ,in ; 2) Parameter initialization, initialize convergence factor Attenuation constant Dynamically adjust parameters Dynamically adjust parameters Dynamically adjust parameters Initialize Gaussian mutation parameters, mean Standard deviation Initialize the positions and fitness values ​​of Alpha, Beta, and Delta for three individual gray wolves; set the acceptance probability of Gaussian mutation local search. ; Fitness assessment for each individual An LSTM neural network model is trained using its representative LSTM hyperparameter configuration, and prediction is performed using engine noise data processed by Empirical Mode Decomposition (EMD). The mean square error (MSE) between the prediction result and the actual noise data is calculated as the fitness value. : in, It is real noise data. It is the prediction result of the LSTM neural network model. n It refers to the number of data points; Location updates include: 1) Calculate the coefficient vector and For each individual Generate random numbers ; Calculate the convergence factor of the nonlinear decay: Calculate the coefficient vector: ， 2) Calculate the dynamic adjustment factor based on the population fitness and the number of iterations: 3) Gray wolf individual position update, for each individual According to its current location Alpha's position Beta position The position of Delta and coefficient vector , and dynamic adjustment factor Update individual location: 4) Gaussian mutation local search, for the updated individual Each dimension Apply Gaussian mutation: in, Represents a standard normal distribution; For the newly generated candidate solutions Fitness evaluation was conducted to obtain ; Based on the set acceptance probability Update individual location: in, r The number is randomly generated. ; Boundary handling: Let the range of values ​​for the hyperparameters of the LSTM neural network model be... ,in and They are the first j The lower and upper bounds of each hyperparameter need to be determined; after each position update, the updated individual position needs to be... Perform boundary checks and processing; For each dimension Perform the following operations: Realize the individual position after boundary processing Satisfy the range constraints of all hyperparameter values; The iteration termination condition is to repeatedly perform fitness evaluation, position update, and boundary handling until the maximum number of iterations is reached. ; The optimal solution is output, which determines the optimal individual position based on the fitness value. and its fitness value ,use The hyperparameter configuration of the representative LSTM neural network model is used to train the final EMD-EGMGWO-LSTM model.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the noise control method for a micro-engine as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the noise control method for a micro-engine as described in any one of claims 1-5.

Images