A method for selecting a reference signal for an active control system of road noise in a vehicle

Abstract

The application belongs to the technical field of automobile, and particularly relates to a selection method of a reference signal of an automobile road noise active control system, which comprises the following steps: 1, constructing a sound pressure level weighted fitness function expression; 2, establishing a mathematical model and setting parameters; 3, determining a gene coding method of a genetic algorithm; 4, generating an initial population of the genetic algorithm; 5, calculating the fitness of each individual in the population; 6, recording the coding and fitness of the optimal individual in the population; 7, designing a child individual according to the actual mathematical model; 8, judging whether the algorithm meets a termination condition, and if not, returning to steps 4-6, and if yes, executing step 9; and 9, outputting the optimal coding recorded, and performing gene decoding and effect verification. The improved genetic algorithm can not only select a reference signal combination more efficiently, but also select a more optimal reference signal combination, so that the noise reduction effect of the active control algorithm is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of automobiles, in particular to a selection method of reference signals of an automobile road noise active control system. BACKGROUND

[0002] With the continuous development of automobile technology, the NVH performance of vehicles has become one of the important indicators for evaluating the comfort of automobiles, and major automobile research centers around the world have invested a lot of manpower and resources in the research and development of low-noise automobiles. At present, the development of new models is often based on old models, and the focus of acoustic design is usually on reducing sound pressure level, and the methods adopted are to increase the stiffness and mass of the driver's cabin panel, and to lay passive noise reduction measures such as sound-absorbing, sound-insulating and damping materials. However, passive noise control has many fatal shortcomings, such as easy to cause hardware to be too heavy, cost increase, and more importantly, passive noise control technology has little effect on low-frequency noise. Compared with passive noise reduction technology, active noise reduction can effectively control the low-frequency noise component by reasonably arranging microphones and secondary sound sources in the scene to eliminate the interference of noise sources, without the need for topology optimization to modify the structure of the target object or the need to set up large-thickness and high-cost sound-absorbing materials.

[0003] Road noise active control is one of the important categories, and usually uses feedforward or feedforward and feedback combined control algorithm, and the control effect is good or bad, which is largely determined by whether the input reference signal is appropriate. At the same time, due to the limitation of hardware computing power and reaction speed, people hope to obtain a reference signal selection scheme that contains the most relevant information and has as few reference signals as possible. However, most methods, such as TPA, weight method, etc., often consume time and effort or require complex experimental conditions, and the efficiency is very low. The proposed reference signal selection method for automobile road noise active control system based on improved genetic algorithm can greatly improve the selection efficiency of reference signals, and only basic data acquisition equipment is needed, which is of great significance to the research of the entire road noise active control system. SUMMARY

[0004] The application provides a selection method of reference signals of an automobile road noise active control system, which uses an improved genetic algorithm to more efficiently select reference signal combinations and to select better reference signal combinations, thereby improving the noise reduction effect of the active control algorithm and solving the above problems existing in road noise control.

[0005] The technical scheme of the application is described below in combination with the drawings:

[0006] A selection method of reference signals of an automobile road noise active control system, comprising the following steps:

[0007] Step 1: Collect data and construct a sound pressure level weighted fitness function expression;

[0008] Step two, establish a mathematical model, and set the parameters initialization;

[0009] Step three, according to the mathematical model to determine the genetic algorithm gene encoding method;

[0010] Step four, according to the constraint conditions to generate the initial population of genetic algorithm;

[0011] Step five, through the sound pressure level weighted fitness function given in step one to calculate the fitness of each individual in the population;

[0012] Step six, record the binary sequence of the optimal individual in this generation and the corresponding fitness value;

[0013] Step seven, according to the actual mathematical model design recombination and mutation operator, through selection, recombination, mutation operator to produce offspring individuals;

[0014] Step eight, determine whether the number of iterations meets the preset maximum iteration limit, if not, return to execute steps four to six, if yes, execute step nine;

[0015] Step nine, output the optimal set of encoding in the record, decode the gene, and input the result into the active noise control system for effect verification.

[0016] Further, the specific method of step one is as follows:

[0017] 11) perform real vehicle data acquisition experiment, obtain N groups of vibration acceleration reference signals collected by several acceleration sensors and M groups of vehicle interior noise signals picked up by microphones;

[0018] 12) establish a sound pressure level weighted fitness function expression according to the data collected in step 11), the sound pressure level weighted fitness function expression is as follows:

[0019]

[0020] In the formula, n f is the discrete number of frequency when solving the coherence coefficient; γ dMxi is the coherence coefficient of the i-th group of reference signals relative to the M-th group of noise signals; SPL M (f) is the sound pressure level of the M-th group of noise signals at frequency f; f min to f max is the selected noise reduction frequency band.

[0021] Further, M≥1 in step 11).

[0022] Further, the specific method of step two is as follows:

[0023] 21) constructing a mathematical model;

[0024] Based on the constructed sound pressure level weighted fitness function, a target function is constructed to maximize the weighted average multiple coherence coefficient of the reference signal and the noise signal, i.e., to maximize the fitness function as the target function;

[0025] The constraint condition of the target function is as follows:

[0026] x j = 0 or 1 j = 1, 2,..., N

[0027] In the formula, x j is the gene code of the jth reference signal, and 1 represents selection and 0 represents non-selection; j takes a value from 1 to N;

[0028] When the number of selected input reference signals is Num, the number of gene codes with 1 in the population individual in the optimization process needs to be Num, so there is a constraint condition as follows:

[0029]

[0030] 22) initializing and setting parameters;

[0031] The parameters in the improved genetic algorithm are initialized, including population size n, gene sequence length length, selected reference signal number Num, iteration number Gmax, recombination probability pc, mutation probability pm, and generation gap GGAP.

[0032] Further, the specific method of step three is as follows:

[0033] The gene length is N, i.e., the number of all collected reference signals, wherein each gene can only be 0 or 1, 0 represents that the chromosome does not include the gene, i.e., the combination does not include the jth reference signal, and 1 represents that the combination includes the jth reference signal.

[0034] Further, the specific method of step four is as follows:

[0035] The individuals in the initial population are generated with a random probability, each individual is a 01 binary sequence with a length of N, and the number of codes with 1 is Num.

[0036] Further, the specific method of step seven,

[0037] The roulette wheel selection operator is used to select individuals participating in the next generation of reproduction, i.e., recombination and mutation, according to the probability, wherein the probability is related to the fitness value of the individual, as shown in the following formula:

[0038]

[0039] In the formula, P(yi ) is the probability of selecting individual y i ; fitness(y i ) is the fitness function value of individual y i , wherein i=1, 2, …, n.

[0040] Further, the step seven,

[0041] The recombination operator, i.e. the crossover operator, adopts the following manner:

[0042] (1) randomly generating 2 positions, and exchanging the sequences in the range of the selected 2 chromosomes;

[0043] (2) counting how many 1 / 0 are changed after the exchange of the two chromosomes;

[0044] (3) in the remaining part of the range, transforming 1 / 0 according to the random selection principle, so as to offset the change and make the number of 0 / 1 codes unchanged

[0045] Further, the step seven,

[0046] The mutation operator adopts the following manner:

[0047] (1) randomly selecting a position in a chromosome coded as 1, and changing 1 to 0;

[0048] (2) randomly selecting a position in a chromosome coded as 0, and changing 0 to 1.

[0049] The beneficial effects of the present application are:

[0050] 1) The present application greatly reduces the time consumption and improves the efficiency;

[0051] 2) The sound pressure level factor of the in-vehicle noise signal is introduced in the fitness function, which enhances the correlation information at the main frequency band of the noise signal;

[0052] 3) The optimal reference signal combination selected by the present application has stronger correlation information, and can improve the noise reduction effect of the active control algorithm when applied to the active noise control technology. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0054] Figure 1 It is a flowchart of the present application;

[0055] Figure 2 To improve the convergence of genetic algorithm;

[0056] Figure 3 The 13-way reference signal selected for the output result and the multiple coherence of all 24-way reference signals relative to the noise signal are compared in the graph;

[0057] Figure 4 The average noise reduction amount graph for the first pickup microphone position of the active control system;

[0058] Figure 5 The average noise reduction amount graph for the second pickup microphone position of the active control system. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0060] Embodiment one

[0061] Referring to Figure 1 A method for selecting reference signals of an active control system for road noise of an automobile, comprising the following steps:

[0062] Step one, collecting data and constructing a sound pressure level weighted fitness function expression;

[0063] 11) Conducting a real vehicle data collection experiment to obtain N groups of vibration acceleration reference signals collected by a plurality of acceleration sensors and M groups of in-vehicle noise signals picked up by microphones;

[0064] 12) Constructing a sound pressure level weighted fitness function expression according to the data collected in step 11), and the sound pressure level weighted fitness function expression is as follows:

[0065]

[0066] In the formula, n f is the discrete number of frequency when solving the coherence coefficient; is the coherence coefficient of the i-th group of reference signals relative to the M-th group of noise signals; SPL M (f) is the sound pressure level value of the M-th group of noise signals at frequency f; f min to f max is the selected noise reduction frequency band.

[0067] Wherein, M >= 1, not only limited to single-channel active noise control system, also applicable to multi-channel active noise control system.

[0068] Step two, establish a mathematical model, parameter initialization settings;

[0069] 21) construct a mathematical model;

[0070] Based on the constructed sound pressure level weighted fitness function, the objective function is constructed to maximize the weighted average multiple coherence coefficient of the reference signal and the noise signal, that is, to maximize the fitness function as the objective function;

[0071] The constraint condition of the objective function is as follows:

[0072] x j = 0 or 1 j = 1, 2,..., N

[0073] In the formula, x j The jth reference signal gene coding, take 1 to represent the selection, take 0 to represent not to select; J takes value from 1 to N;

[0074] When the number of selected input reference signal is Num, the number of gene coding with 1 in the population individual in the optimization process is Num, so there is a constraint condition as follows:

[0075]

[0076] 22) parameter initialization settings;

[0077] The initialization of the parameters in the improved genetic algorithm includes: population size n, gene sequence length length, selected reference signal number Num, maximum iteration number Gmax, recombination probability pc, mutation probability pm, generation gap GGAP.

[0078] Step three, determine the genetic coding method of genetic algorithm according to the mathematical model;

[0079] The gene length is N, that is, the number of all reference signals collected, wherein each gene can only be 0 or 1, 0 represents that the chromosome does not include the gene, that is, the combination does not include the jth reference signal, and 1 represents that the combination includes the jth reference signal.

[0080] Step four, generate the initial population of genetic algorithm according to the constraint condition;

[0081] The individual in the initial population is generated with random probability, each individual is a 01 binary sequence with length N, and the number of coding with 1 is Num.

[0082] Step five, calculate the fitness of each individual in the population through the sound pressure level weighted fitness function given in step one;

[0083] Step six, record the binary sequence of the individual with the optimal fitness in the population of this generation and its corresponding fitness value;

[0084] Step seven, design recombination and mutation operators according to the actual mathematical model, and generate offspring individuals through selection, recombination and mutation operators;

[0085] The selection operator adopts a roulette wheel operator, and individuals participating in the next generation of reproduction, i.e. recombination and mutation, are selected according to a probability, wherein the probability is related to the fitness value of the individual, and is shown in the following formula:

[0086]

[0087] In the formula, P(y i ) is the probability of selecting individual y i ; fitness(y i ) is the fitness function value of individual y i , wherein i = 1, 2, …, n.

[0088] The recombination operator, i.e. the crossover operator, adopts the following manner:

[0089] (1) randomly generate two positions, and exchange the sequences in the range of the selected two chromosomes;

[0090] (2) count how many 1 / 0s are changed in the two chromosomes after the exchange;

[0091] (3) in the remaining part of the range, transform 1 / 0 according to the random selection principle, so as to offset the changes and keep the number of 0 / 1 unchanged;

[0092] The mutation operator adopts the following manner:

[0093] (1) randomly select a position in a chromosome coded as 1, and change 1 to 0;

[0094] (2) randomly select a position in a chromosome coded as 0, and change 0 to 1.

[0095] Step eight, judge whether the iteration number meets the preset maximum iteration number limit, if not, return to execute steps four to six, and if yes, execute step nine;

[0096] Step nine, output the optimal set of codes in the record, perform gene decoding, and input the result into the active noise control system for effect verification.

[0097] Example two

[0098] This example takes a certain vehicle as an example to introduce the reference signal selection method of the automobile road noise active control method based on the improved genetic algorithm, which is as follows:

[0099] Step one, when carrying out the real vehicle data acquisition experiment, 8 acceleration sensors are used to collect 24 groups of reference signals, and 2 groups of in-vehicle noise signals are picked up by microphones, and the microphones are placed at the headrests of the main driver and the co-driver respectively;

[0100] The fitness function constructed is:

[0101]

[0102] Wherein, n f is the discrete number of frequency when solving the coherence coefficient; SPL M (f) is the sound pressure level of the Mth group of noise signals at frequency f; is the coherence coefficient of the ith group of reference signals relative to the Mth group of noise signals; f min to f max is the selected noise reduction frequency band, and the embodiment selects 25-500 noise reduction frequency band.

[0103] The calculation formula of in the fitness function is as follows:

[0104]

[0105] Wherein, is the cross power spectral density between the ith group of reference signals and the Mth group of in-vehicle noise signals; are the self-power spectral densities of the ith group of reference signals and the Mth group of in-vehicle noise signals respectively.

[0106] Step two, the mathematical model constructed is as follows:

[0107] Based on the constructed fitness function, a target function is constructed to maximize the weighted average multiple coherence coefficient of the reference signal and the noise signal, that is, to maximize the fitness function as the target function of the application;

[0108] The constraint condition of the target function is as follows:

[0109] x j =0 or 1 j=1,2,...,N

[0110] In the formula, x j is the gene code of the jth reference signal, which takes 1 to represent selection and 0 to represent non-selection; j takes a value from 1 to N.

[0111] When the number of selected input reference signals is Num, it is necessary to ensure that the number of gene codes taking 1 in the population individuals in the optimization process is Num, so the constraint condition is as follows:

[0112]

[0113] The parameters in the improved genetic algorithm are initialized, including population size n=50, gene sequence length length=24, selected reference signal quantity Num=13, iteration number Gmax=50, recombination probability pc=0.9, mutation probability pm=0.1, and generation gap GGAP=0.9.

[0114] In step three, the gene coding method of the improved genetic algorithm is as follows: the gene length is N, that is, the quantity of all collected reference signals is 24, wherein each gene can only be 0 or 1, 0 represents that the chromosome does not include the gene, that is, the reference signal is not included in the combination, and 1 represents that the reference signal is included.

[0115] In step four, the initial population of the genetic algorithm is generated according to the constraint condition, the individual in the initial population is generated in a random probability, each individual is a 01 binary sequence with a length of N, and the quantity of codes with 1 is Num.

[0116] In step five, the sound pressure level weighted fitness function given in step one is used to calculate the fitness of each individual in the population.

[0117] In step six, the binary sequence of the individual with the optimal fitness in the population and the corresponding fitness value are recorded.

[0118] In step seven, the recombination and mutation operators are designed according to the actual mathematical model, and the offspring individuals are generated through the selection, recombination and mutation operators.

[0119] The roulette wheel selection operator is used to select the individuals participating in the next generation breeding according to the probability, wherein the probability is related to the fitness value of the individual, and the probability is shown in the following formula:

[0120]

[0121] In the formula, P(y i ) is the probability of selecting the individual y i ; fitness(y i ) is the fitness function value of the individual y i , wherein i=1, 2, …, n.

[0122] The recombination operator, that is, the crossover operator, is used in the following manner:

[0123] (1) two positions are randomly generated, and the sequences in the range of the selected two chromosomes are exchanged;

[0124] (2) the number of 1 / 0 changed in the two chromosomes after the exchange is counted;

[0125] (3) in the remaining part of the range, the 1 / 0 is transformed according to the random selection principle, so as to offset the change and keep the quantity of codes 0 / 1 unchanged.

[0126] The mutation operator adopts the following manner:

[0127] (1) randomly selecting a position coded as 1 in a chromosome, changing 1 to 0;

[0128] (2) randomly selecting a position coded as 0 in a chromosome, changing 0 to 1.

[0129] Step eight, judging whether the iteration number meets the preset maximum iteration number limit, if not, returning to execute steps four to six, and if yes, executing step nine;

[0130] Step nine, outputting the optimal set of codes in the record, performing gene decoding, and inputting the result into the active noise control system for effect verification.

[0131] The convergence process of the algorithm in this embodiment is shown in Figure 2 , and it can be seen that the algorithm converges at about 20 generations. The average multiplicity correlation coefficient of the 13 reference signals obtained is Figure 3 . As can be seen from the figure, when the number of reference signal channels is reduced by nearly half, the multiplicity at each frequency point does not decrease much; inputting the result into the active noise control system, the system uses a double-channel feedforward NFxLMS algorithm, uses two secondary loudspeakers and two error microphones, and obtains the average noise reduction amount of the first channel and the second channel, and the result is shown in Figures 4-5 .

[0132] In summary, it is found that the average noise reduction amount of the first channel can reach 6 dB, and the average noise reduction amount of the second channel can reach 5 dB, and the present application can achieve good noise reduction effect.

[0133] The preferred embodiments of the present application are described in detail above with reference to the accompanying drawings, but the protection scope of the present application is not limited to the specific details in the above-described embodiments. Any person skilled in the art can make equivalent substitutions or changes to the technical solutions and the inventive concept of the present application within the technical scope disclosed by the present application, and these simple modifications all belong to the protection scope of the present application.

[0134] In addition, it should be noted that each specific technical feature described in the above specific embodiments can be combined in any appropriate manner without contradiction, and in order to avoid unnecessary repetition, the present application will not further describe various possible combinations.

[0135] In addition, various different embodiments of the present application can also be combined in any manner, as long as they do not violate the idea of the present application, and they should also be considered as disclosed by the present application.

Claims

1. A method for selecting a reference signal for an active control system for road noise in a vehicle, characterized in that The method comprises the following steps: Step one, collect data and construct a sound pressure level weighted fitness function expression; Step two, establish a mathematical model and set the parameters; Step three, determine the gene coding method of the genetic algorithm according to the mathematical model; Step four, generate the initial population of the genetic algorithm according to the constraint conditions; Step five, calculate the fitness of each individual in the population by the sound pressure level weighted fitness function given in step one; Step six, record the binary sequence of the optimal individual in the population and the corresponding fitness value; Step seven, design recombination and mutation operators according to the actual mathematical model, and generate offspring individuals through selection, recombination and mutation operators; Step eight, determine whether the iteration number meets the preset maximum iteration number limit, if not, return to steps four to six, and if yes, execute step nine; Step nine, output the optimal set of codes recorded, decode the genes, and input the results into the active noise control system for effect verification; The specific method of step one is as follows: 11) Perform a real vehicle data collection experiment to obtain N groups of vibration acceleration reference signals collected by a plurality of acceleration sensors and M groups of vehicle interior noise signals picked up by a microphone, M≥1; 12) Construct a sound pressure level weighted fitness function expression according to the data collected in step 11), and the sound pressure level weighted fitness function expression is as follows: where n f to solve the discrete number of frequencies of the coherence coefficient; is the coherence coefficient of the ith set of reference signals with respect to the Mth set of noise signals, where i = 1, 2, …, n; SPL M (f) is the sound pressure level of the Mth set of noise signals at frequency f; f min to f max is the selected noise reduction frequency band; In the fitness function The formula for calculating the following: wherein, is the cross power spectral density between the i-th group of reference signals and the M-th group of in-vehicle noise signals; is the auto power spectral density of the i-th group of reference signals and the M-th group of in-vehicle noise signals, respectively. The specific method of step two is as follows: 21) Construct a mathematical model; Based on the constructed sound pressure level weighted fitness function, construct a target function to maximize the weighted average multiple coherence coefficient of the reference signal and the noise signal, that is, maximize the fitness function as the target function; The constraint conditions of the target function are as follows: x j = 0 or 1 ; j = 1, 2,..., N In the formula, x j is the gene code of the jth reference signal, and 1 represents selection and 0 represents non-selection; j takes a value from 1 to N; After selecting the number of input reference signals as Num, the number of gene codes with value 1 in the population individuals in the optimization process is Num, so there is the following constraint condition: 22) Initialize the parameters; The parameters in the improved genetic algorithm include: population size n, gene sequence length length, selected reference signal number Num, iteration number Gmax, recombination probability pc, mutation probability pm, and generation gap GGAP.

2. The method for selecting a reference signal for an active vehicle road noise control system according to claim 1, characterized in that, The specific method of step three is as follows: The gene length is N, that is, the number of all collected reference signals, wherein each gene can only be 0 or 1, 0 represents that the chromosome does not include the gene, that is, the reference signal is not included in the combination, and 1 represents that the reference signal is included.

3. The method for selecting a reference signal for an active vehicle road noise control system according to claim 1, characterized in that, The specific method of step four is as follows: The individuals in the initial population are generated with a random probability, each individual is a 01 binary sequence with a length of N, and the number of codes with value 1 is Num.

4. The method for selecting a reference signal for an active vehicle road noise control system according to claim 1, characterized in that, The specific method of step seven is as follows: The selection operator adopts a roulette wheel operator, and individuals participating in the next generation of reproduction, that is, recombination and mutation, are selected according to the probability, wherein the probability is related to the fitness value of the individual, as shown in the following formula: where P(y i ) is the probability of selecting individual y i ; and fitness(y i ) is the fitness function value of individual y i , where i = 1, 2, …, n.

5. The method for selecting a reference signal for an active vehicle road noise control system according to claim 1, characterized in that, The specific method of step seven is as follows: The recombination operator, that is, the crossover operator, adopts the following method: (1) Randomly generate two positions, and exchange the sequences in the selected two chromosomes; (2) Count how many 1 / 0s are changed in the two chromosomes after exchange; (3) In the remaining part of the range, transform 1 / 0 according to the random selection principle, so as to offset the change so that the number of encoded 0 / 1 is unchanged.

6. The method for selecting a reference signal for an active vehicle road noise control system according to claim 1, characterized in that, Said step seven, The mutation operator adopts the following way: (1) Randomly select a position in the chromosome coded as 1, change 1 to 0; (2) Randomly select a position in the chromosome coded as 0, change 0 to 1.

Images