Research On Multi-channel Adaptive Active Noise Control Systems For Reducing Vehicle Interior Low Frequency Noise

The interior noise in automotive cabin has been one of the importantindexes used to evaluate the ride comfort performance, as a result of theincreasing requirements of automotive performances. Now, the commonpassive reduction noise measures in automobile, such as noise elimination,sound absorption, vibration and sound isolation, etc, have good effect onreducing the middle and high frequency noise. But when using theabove-mentioned measures to eliminate the vehicle interior noise, thedifferent methods must be applied on different types of vehicle, and thesemeasures also be expensive and have little effect on low frequency. Whilethe adaptive active noise control(AANC)is very effective on reducing lowfrequency, and AANC has the less bulk , lighter weight than the passiveacoustic measures, and it is prone to control. Along with the development ofthe modern control technology and descend of the COMS chip's price,AANC has more and more advantage than the common passive measures. Therefore, a vehicle interior adaptive active noise control system ofmulti-channel is researched in this paper. A vehicle interior adaptive activenoise control method of multi-channel based on online training dynamicneural network model is proposed, and the AANC system of multi-channelis established based on this method. The corresponding adaptive activenoise control program code is developed, and vehicle interior adaptiveactive noise control experiments are carried out. This research hasprovided a practicality base on applied work of the vehicle interior adaptiveactive noise control system. The main research works completed in thispaper are as follows: －4－吉林大学博士学位论文 After summarizing the feedforward active noise control system, aviewpoint that taking the DNN method as the key arithmetic of the activenoise control is put forward. Based on the fastest declined theoretics, theweight modification of dynamic neural network method is deducedparticularly, and the numeric expression between the feedback layer outputand hide layer is set up, which improves the operate precision. The inputsignals of DNN are the real measured 5 kinds vibration acceleration signalsand the being-distinguished signals are the noise signals beside the drivers'and co-pilots' ear, Every parameter of the DNN has been optimized withapplying the Matlab/Simulink toolbox. And making use of optimized nervenetwork structure, the interior noise in automotive cabin is distinguished ondifferent rotate speeds. The results indicate that the optimized network hassome excellences , such as simplified-structure, and reasonableparameter-matching, which can achieve the pre-desire of the paper . The model of vehicle interior multi-channel adaptive active noisecontrol system is set up and the control strategy is put forward. Based onthe control strategy of the above-mentioned model, the Multi-channel DNNarithmetic—MDNN is put forward, which suit to vehicle interiormulti-channel adaptive active noise control. Then the sound path problem inmulti-channel adaptive active noise control system is studied with thisarithmetic, and a method in which this situation can be fit by counteractingpath network is proposed. Due to this dynamic neural network structurebeing convenient for real-time application, the referent-signal andcounteract path can be distinguished in real-time by the arithmetic, and thereal online-practice of dynamic neural network can be carried out, whichimproves the ability of suiting to the situation alteration in driving, and hasthe strong suit-capability and practical- capability. Using Matlab/Simulink toolbox, based on the arithmetic MDNN, themode of a vehicle interior adaptive active noise control system ofdual-channel is designed, and utilizing the testing points' signals ofacceleration of vibration and the two noise signals in vehicle, the activenoise control system is simulated. The results of s...