4.1 Passive silencer optimized for AENC
One difficulty experienced was that the exhaust noise in the high frequency range was reduced by the active noise cancellation system. Therefore, the passive silencer, optimized for high frequency noise reduction, and the active silencer were combined. The second role of this passive silencer was that the sound field of the active silencer part was separated from the sound field found upstream. Generally, the distance from the engine to AENC should be different according to the condition of installation. In this way, the optimization of the shape in the active silencer area could work appropriately.
In case of AENC for the anxiously engine, in order to improve problems, such as that the sound emitted by the secondary speaker being de-amplified at the second order of the resonance frequency, the Helmholz resonator was assembled in the passive silencer. As a result, the second order level of the exhaust noise was decreased to less than the noise level on which the secondary speaker could emit.
4.2 Acoustic Analysis in Whole System
In AENC for auxiliary engine specifications, the acoustic behavior of the whole system could be analyzed by SYSNOISE in Fig.13. Fig.13a) shows the distribution of sound pressure at high frequency range. The noise can be reduced by the passive silencer. But it can reduce the noise in the low frequency range, such as the first order of exhaust noise, by only a little as in fig.13b). The reason for this was that the absorption type silencer was used to reduce high frequency noise.
The results of the analysis of AENC is shown in Fig.13c). In the upstream of the acoustic mixing area, the sound pressure of the first order of exhaust noise is still at a high level. But beyond the mixing area, it is clear that the exhaust noise could be reduced drastically.
5. Optimization of Controller Algorithm
5.1 Adaptive Controller
The algorithm of an adaptive control is shown in figure 14. The output y(n) to the speaker in time n was calculated by convoluting the input x(n) of the reference microphone and the coefficient of the adaptive filter W as follows:
The speaker was driven by y(n). Next, the secondary noise at the error microphone could be calculated by convolution of the acoustic transfer function C. Also, the exhaust noise at the same point could be calculated by convolution of the acoustic transfer function of system T and noise x(n). Therefore, the remaining noise ε(n) was indicated by:
ε(n) ＝ T (n)x(n) ＋ C (n)y(n) ... (4)
In order to minimize the power of the error, the coefficient of the adaptive filter was updated by Filtered-X Algorithm, such as:
