† Corresponding author. E-mail:
Project supported by the National Natural Science Foundation of China (Grant Nos. 11574249 and 11874303) and the Natural Science Basic Research Plan in Shaanxi Province of China (Grant No. 2018JQ1001).
The efficient and accurate synthesis of physical parameter-controllable impact sounds is essential for sound source identification. In this study, an impact sound synthesis model of a cylinder is proposed based on discrete state space (DSS) method and modal extension method (MEM). This model is comprised of the whole three processes of the physical interaction, i.e., the Hertz contact process, the transient structural response process, and the sound radiation process. Firstly, the modal expanded DSS equations of the contact system are constructed and the transient structural response of the cylinder is obtained. Then the impact sound of the cylinder is acquired using improved discrete Raleigh integral. Finally, the proposed model is verified by comparing with existing models. The results show that the proposed impact sound synthesis model is more accurate and efficient than the existing methods and easy to be extended to the impact sound synthesis of other structures.
Radiated sounds in our daily life usually vary with the physical parameters of vibrating structures. These physical parameters include size, shape, and material, as well as stimuli and radiation condition.[1] Investigating the relationships between physical parameters of structures and sound features is meaningful and attractive in the field of sound source identification.[2] In order to grasp the most essential features of the vibratory and acoustical problem, sound continua should be created since they can simulate the continuous changes of the physical parameters.[3] Recordings or real sounds are not efficient to create continua since actual physical parameters are in inhomogeneous and disperse distribution.[4,5] By contrast, synthetic sounds, which can precisely control different physical parameters and provide large sets of sound samples, seem more suitable. In previous literatures, the synthesized impact sounds are mainly used in sound source identification studies since they are brodband and carrying most of the information related to physical parameters.[6,7] Thus, the synthesis of impact sound continua becomes the premised issue in the study of sound source identification.
A complete impact sound synthesis model should comprise three processes, the contact process of the hammer and the structure, the vibration response process of the structure, and the sound radiation process. The contact process was substituted by pre-defined contact forces in wave forms of a half sine or two half-Gaussians which rise fast and decay slow in some studies.[5,8] Meanwhile, the other studies used the Hertz contact model, which is closer to the reality compared with pre-defined forces, to stimulate the whole processes of the contact.[9,10] The vibration response and radiation process was usually modeled with basic theoretical derivation, modal expansion method (MEM), finite difference method (FDM), finite element method (FEM), boundary element method (BEM), and combination of the above methods.[9–13] Lambourg (2001) synthesized the vibration response and impact sound through the finite-difference time-domain (FDTD) method and discrete Raleigh integral, but his method had a drawback of low computational efficiency.[9] Lutfi (2001) and McAdams (2010) used simplified models based on Lambourg’s method to synthesize the impact sound of bar and plate for subjective judgments, but their models would not be suitable for the accurate-sound-feature extraction since they ignored the information which is not sensitive by the human ears.[10,11] Zhang (2014) proposed FDTD-MEM method and FEM-MEM-BEM method (implemented in LMS Virtual.Lab software) to synthesize impact sound more efficiently than FDTD method under the same precision.[12,13] However, Zhang’s methods still suffered from low computational efficiency caused by high order FDTD and large number of finite elements. In addition, Avanzini (2001) put forward a low parameter sound synthesis model, from which the system was constructed by DSS equations discretized through bilinear transform method.[14] The DSS equations are first order with lower computing cost than high order FDTD equations. Nevertheless, Avanzini’s model neglects both stimulus-position information and radiation process, and the model is not accurate due to the bilinear transform method adopted to discrete the state space equations. Although aforementioned methods have advantages respectively in accuracy and efficiency, an efficient and accurate method for sound synthesis still needs to be developed for sound feature identification tasks.
In order to have a simultaneous improvement on the accuracy and the computational efficiency compared to aforementioned methods and ensure the information integrity of the synthesized sounds, an impact sound synthesis model is proposed in this study. The model comprises the Hertz contact process, the vibration response process, and the sound radiation process. And it is firstly built with modal expanded DSS equations. Then the vibration response and radiated sound of the simply-supported cylinder, which is widely used in airplane and ship manufactures, are calculated using two approximate methods (Euler method and trapezoid method). The results show that the proposed sound synthesis model is superior to the existing methods in both accuracy and efficiency. The proposed model can provide a large number of sound samples for the subsequent sound source identification study and it can be extended to other structures with just a few modifications of the modal parameters.
The impact sound synthesis model used in this paper is described briefly in Fig.
Basic equations of different parts of the sound synthesis model are displayed in the following subsections. They will be discretized and solved in a form of regressive iteration equations in Section
In order to decrease the dimension of iterative computation, the equations of the contact system are usually modal expanded using MEM. The modal expanded second order differential equation of the cylinder can be expressed as follows:
![]() |
![](cpb_27_11_114302/cpb_27_11_114302_ieqn1.gif)
![](cpb_27_11_114302/cpb_27_11_114302_ieqn2.gif)
![](cpb_27_11_114302/cpb_27_11_114302_ieqn3.gif)
![](cpb_27_11_114302/cpb_27_11_114302_ieqn4.gif)
![]() |
![]() |
Assuming the contact process is on the horizontal direction, the equation of hammer motion can be described as
![]() |
Contact force (or impact force) is one of the crucial clues associated with the physical parameters of the hammer, such as the hardness.[16] As mentioned in Section
![]() |
In order to improve the accuracy and the computational efficiency of aforementioned methods, we propose an accurate DSS and MEM based method to calculate the structural response of the cylinder. The modal expanded DSS equation of Eq. (
![]() |
![]() |
![]() |
![]() |
![](cpb_27_11_114302/cpb_27_11_114302_ieqn6.gif)
![](cpb_27_11_114302/cpb_27_11_114302_ieqn7.gif)
![](cpb_27_11_114302/cpb_27_11_114302_ieqn8.gif)
![]() |
![]() |
![]() |
![]() |
![]() |
Accordingly, the DSS equations of the hammer and the contact model can be described as
![]() |
![]() |
![](cpb_27_11_114302/cpb_27_11_114302_ieqn9.gif)
![](cpb_27_11_114302/cpb_27_11_114302_ieqn10.gif)
![](cpb_27_11_114302/cpb_27_11_114302_ieqn11.gif)
By solving Eqs. (
![]() |
The parameters of the contact system are given in Table
![]() | Table 1. Physical parameters of the system. . |
The temporal contact force and velocity response of the cylinder from different approximate methods are displayed in Figs.
The FEA software can hardly work out a result under the calculation precision of 5000 Hz since the required element number for the cylinder in Table
![]() |
![]() |
The significant deviation of Avanzini's bilinear transform result can be interpreted from two aspects: 1) the Duhamel integral in Eq. (
As a complementary validation, the velocity response of a smaller cylinder (L = 0.3 R=0.1) calculated using FEM+MEM (realized through the transient modal superposition structural response case in Virtual.Lab software) and proposed method is shown in Fig.
The time consumed in different steps of the transient structural response process is displayed in Table
![]() | Table 2. Time consumed in different steps. . |
It is noticed that the proposed methods, especially the trapezoid approximate method, are more efficient than the bilinear transform method used by Avanzini. This phenomenon reveals that pre-calculation of two terms in Eq. (
A classical way to calculate the sound field of a cylinder is the continuous Raleigh integral method in cylindrical coordinate,[15] but it is too complex to meet the requirement of high computational efficiency. On the contrary, discrete Raleigh integral is more efficient since the computational dimension is greatly reduced by the discretization operation.[9,13] There are two discrete forms of Raleigh integrals in Eq. (
![]() |
![]() |
![](cpb_27_11_114302/cpb_27_11_114302_ieqn12.gif)
![](cpb_27_11_114302/cpb_27_11_114302_ieqn13.gif)
![]() |
The receiving point of the sound field is set to xrecieve(r, θ,z) = (2,0,0.08) and the impact sounds of different Raleigh integral forms are shown in Fig.
It is obvious that the impact sounds from different methods well agree with each other. The proposed first order method has the same accuracy to the second order FDTD method used by Lambourg and Zhang.
For further validation of the proposed impact sound synthesis method, the impact sound of a smaller cylinder (L = 0.3 R = 0.1) calculated respectively using FEM+MEM+BEM (realized through the transient modal superposition weakly coupled response case in Virtual.Lab software) and proposed method are shown in Fig.
The computing time of the impact sound (does not contain the Hertz contact and velocity response processes) is displayed in Table
![]() | Table 3. Comparison of the impact sound computation efficiency. . |
Based on a simply supported cylindrical shell structure, an impact sound synthesis model comprising the Hertz contact process, the transient structural response process, and the sound radiation process is proposed. Firstly, the system equations are constructed through the modal extended DSS equations and the transient structural responses of the cylinder are obtained. Then the impact sound is calculated using the proposed first order discrete Rayleigh integral method. Finally, the sound synthesis model is verified in two steps: 1) verifying the structural response via the point admittance and FEM+MEM method, and 2) verifying the sound radiation by comparing impact sounds from different methods.
The results show that the DSS and MEM based sound synthesis model has advantages over the existing sound synthesis models in terms of both computational efficiency and accuracy. The sound synthesis model could provide a large number of effective sound samples for sound source identification study and could be easily transferred to the impact sound synthesis of other structures.