Professional Documents
Culture Documents
Computational Fluid Dynamics On Sounding
Computational Fluid Dynamics On Sounding
Computational Fluid Dynamics On Sounding
375–381 (2008)
* Toyota Motor Corporation, 1 Toyota-cho, Toyota City, Aichi Prefecture 471-8571, Japan
** Prometech Software Inc., 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan
*** Toyo University, 2-36-5 Hakusan, Bunkyo-ku, Tokyo 112-0001, Japan
1-32-19, Akatsutsumi, Setagaya-ku, Tokyo, Japan
E-Mail: yagawa@eng.toyo.ac.jp (Corresponding Author)
ABSTRACT: The sounding mechanism in air-reed instruments is investigated based on the computational fluid
dynamics. While previous studies have assumed a sound source based on the acoustic theory, the authors attempt in
the present study to explain the source of sound as an essential factor defining the timbre of musical instruments. To
deal with the high computational cost required to capture minute changes in pressure, a parallel supercomputer is
employed for computation. The computational results are consistent with Brown’s experimental equation, and new
frequency components revealed only in this three-dimensional analysis are identified and attributed to the harmonics
of the air jet.
Keywords: computational fluid dynamics, finite element method, edge tone, reed instruments, sounding
mechanism
375
Engineering Applications of Computational Fluid Mechanics Vol. 2, No. 3 (2008)
numerical simulation of musical instrument sound Temporally discretizing Eqs. (1) and (2) by the
will lead to the establishment of a scientific fractional step method (Donea, Giuliani and Lavel,
measure for designing instruments (Takeuchi, 1982), we have
Yamamoto and Kawauchi, 1988). Also, this study
can be directly applied to the design of rapid u n +1 − u n 1
+ (u n ⋅ ∇)u n = −∇p n +1 + Δu n (3)
transit railway, car, airplane, electric device, etc., Δt Re
in which reduction of the fluid noise is of primary ∇ ⋅ u n+1 = 0 (4)
concern.
where n is the time step. Then, the intermediate
2. GOVERNING EQUATIONS AND flow velocity u~ is defined as follows,
~ = u n − Δt ⎧(u n ⋅ ∇)u n − 1 Δu n ⎫
SOLUTION ALGORITHM
⎨ ⎬
⎩ ⎭
u (5)
2.1 Finite element fluid analysis and hybrid Re
solution method where Δt is the time increment. By taking the
As sound propagates in a medium as a divergence of Eq. (3) and substituting Eq. (4), we
longitudinal pressure wave, acoustic phenomena obtain the following Poisson’s equation for
have traditionally been regarded as compressible pressure.
fluid mechanics. However, the variation in 1
∇p n+1 = ~
∇⋅u (6)
pressure due to the passage of sound is of the Δt
order of 10-3 to 10-5 times the static pressure of the
ambient fluid, rendering it difficult to capture Then, by subtracting Eq. (5) from Eq. (3), the
both the static pressure of the fluid and the sound following equation for velocity is obtained.
pressure simultaneously with satisfactory ~ − Δt∇p n+1
u n+1 = u (7)
accuracy. Furthermore, the simulation of sound
propagation as a compressible fluid phenomenon The above calculation procedures are summarized
in real-world volumes requires enormous as follows. First, the intermediate flow velocity u~
computational power, and remains almost is calculated using Eq. (5), and pressure pn+1 is
impossible even using today’s supercomputers. calculated by substituting u~ for Eq. (6). Then
Solving low Mach number flow using the velocity un+1 is calculated by substituting the
compressible Navier-Stokes equation numerically pressure for Eq. (7). Since the Reynolds number
also involves a stiffness problem and the of the flow in this study is under 500, the upwind
numerical solution may not converge or accuracy method and the turbulence model are not adopted.
may be severely degraded. Both velocity and pressure fields are interpolated
In this study, a hybrid solution method (Kato et using linear tetrahedral elements (P1-P1 element).
al., 1994), in which sound generation and sound
propagation are analyzed separately, is adopted. 2.3 Lighthill’s theory and Curle’s equation
The behavior of air is computed assuming
incompressible flow using the computed pressure In air-reed instruments, the Mach number of air
change; generated sound is computed by jets around the edge is at most 0.06. Under this
aerodynamic sound prediction based on the condition, if the distance between the sound
Lighthill theory (1952). The finite element source and the observation point is sufficiently
method (FEM) and the fractional step method long, the Curle’s equation (Curle, 1995), a
(Donea, Giuliani and Lavel, 1982) are adopted as solution of the Lighthill’s theory (Lighthill, 1952),
∫ n p(y, t − x / c )dS
the numerical analysis scheme of fluid. can be derived as
1 xi ∂
2.2 Discretization of Navier-Stokes equation pα (x, t ) = (8)
4πc0 x 2 ∂t
i 0
S
The Navier-Stokes equation and continuity
equation assuming negligible external force are as where x is the coordinate of the observation point,
follows. p is the pressure against the object surface, S is
the object surface, x = | x|, y is the coordinate in
∂u 1
+ (u ⋅ ∇) = −∇p + Δu (1) the analysis domain, ni is the outward normal
∂t Re vector on S, and c0 is the speed of sound in the
∇⋅u = 0 (2) static fluid. Therefore, if the pressure of the edge
where u and p are normalized velocity and surface is found, the sound pressure at the
pressure, and Re is the Reynolds number. observation point can be determined using Eq. (8).
376
Engineering Applications of Computational Fluid Mechanics Vol. 2, No. 3 (2008)
377
Engineering Applications of Computational Fluid Mechanics Vol. 2, No. 3 (2008)
Table 2 Analysis conditions of each simulation case. window and no smoothing technique was
employed.
Representative
velocity Reynolds number Analysis time
Case 1 3.0 m/s 195.18 0.067 s (100,000 steps)
Case 2 5.0 m/s 325.3 0.040 s (100,000 steps)
Case 3 7.0 m/s 455.42 0.043 s (150,000 steps)
378
Engineering Applications of Computational Fluid Mechanics Vol. 2, No. 3 (2008)
simulated results in addition to the major frequency component corresponding to Brown’s experimental
equation. These minor components are considered to be the harmonic components of the edge tone.
Fig. 5 Sound pressure vs. time (Case 1). Fig. 6 Power spectrum (Case 1).
Fig. 7 Sound pressure vs. time (Case 2). Fig. 8 Power spectrum (Case 2).
Fig. 9 Sound pressure vs. time (Case 3). Fig. 10 Power spectrum (Case 3).
379
Engineering Applications of Computational Fluid Mechanics Vol. 2, No. 3 (2008)
4. CONCLUSIONS
380
Engineering Applications of Computational Fluid Mechanics Vol. 2, No. 3 (2008)
381