An Inverstigation On The Multi-Objects Optimization Method And Its Experiments For The Configuration Of Random Acoustic Array

As an important acoustic measurement and signal processing method,acoustic array technology is widely used in sound source identification,tracking,monitoring and other aspects.In recent years,the measurement performance of acoustic array is facing severe challenges for the complexity improvement of sound source type and sound field.The acoustic array configuration is the vital factor to improve the measurement performance.However,the traditional array configuration has poor randomness and strong empirical dependence.The complex nonlinear relationship between array geometry and measurement performance cannot be explained in detail,and the reasonable configuration basis and effective criteria cannot be given,resulting in strong blindness of array configuration,high complexity of generation process and poor adaptability.The configuration basis,criteria and methods need to be studied and solved urgently.Therefore,this paper proposes a multi-objective optimization and efficient configuration method for random acoustic arrays,sets reasonable structural constraints,interprets and quantifies the relationship between structure and performance,and gives the basis and criteria for rapid construction of specific arrays.This thesis is supported by the National Natural Science Foundation of China(61871447).The main research work and contents are as follows :(1)Based on the principle of delay-sum beamforming and sparse array configuration,a random array configuration strategy based on regular array and sub-array mode is proposed.The relationship between array factor and beam mode is derived,and the influence of array geometric structure parameters on beamforming is clarified.The key factors such as array aperture,microphone number and partition mode are given,and quantified,and the quantitative standard is formulated.(2)The construction strategy of coaxial ring random array is proposed.The uniformity condition of microphone distribution is set.The array partition,microphone position and beam performance are optimized with the uniformity index as the constraint condition.In the polar coordinate system,the mathematical model of array microphone position is constructed.The complex and lengthy microphone position search is converted into the mathematical construction process of position matrix,which greatly reduces the target array generation time.(3)Based on the nonlinear multi-objective correspondence between array structure and measurement performance,a multi-objective optimization configuration method of random acoustic array is constructed.The Non-dominated Sorting Genetic Algorithm II is used to construct the multi-objective optimization configuration model of the array by combining the main lobe width and sidelobe suppression ratio of the array performance evaluation parameters.The NSGA2 algorithm is improved for the premature convergence or local convergence problem of the NSGA2 algorithm and the actual demand of the array configuration.(4)The effectiveness of the method and its key algorithms is verified by numerical simulation and laboratory and field experiments.The array structure,performance and their relationship are analyzed by statistical analysis.The efficiency of the method is verified from the perspective of sound source identification through field experiments.Simulation and experimental results show that this method can not only quickly and efficiently construct a random array according to the type of sound source,but also has good adaptability and robustness to complex acoustic environment.