Sunteți pe pagina 1din 7

2003-01-1625

Virtual Design of Brake Squeal


Chih-Hung Jerry Chung William Steed Jianrong Dong
MTS SYSTEMS

Bong Soo Kim Geun Soo Ryu


Hyundai Motor Co.
Copyright 2003 SAE International

INTRODUCTION
Disk Brake Squeal noise is a problem that continues to confront automobile manufacturers. Customer complaints result in significant warranty costs yearly. Furthermore, customer dissatisfaction can cause a loss of future business. Physically, squeal noise occurs when the friction coupling between the rotor and pad creates a dynamic instability. This leads to vibration of the structure, which radiates a high frequency noise in the 115 kHz range. Many analytical approaches have been proposed in the literature to evaluate the Brake Squeal Dynamics and the most popular approach is the Complex Eigenvalue approach[3]-[6]. Root Locus analysis is a further application of the Complex Eigenvalue approach, which can trace the unstable modes back to meaningful structure mode pairs. While the Root Locus method provides a way to identify potential critical modes, it suffers from several drawbacks, such as long solution time and lack of resolution to provide the detail mode interaction among complex roots. Chung, et. al. [1] presented a new analysis approach by transferring the brake system equation of motion from transient domain to modal domain. The modal domain transformation significantly reduces the complexity of the Complex Eigenvalue analysis and provides mechanism of mode coupling phenomenon. The presented approach has been successfully applied to solve Automotive Brake Squeal problem [1] and Motorcycle brake squeal problem [2]. The goal of this paper is to apply the Modal Domain analysis approach to minimize design iterations using FE models and provides a new method to calculate Complex Eigenvalues to reduce solution time.

The title of the paper is called Virtual Design of Brake Squeal. Industry usually refers the Virtual Design as using models to perform design studies before really changing the hardware. The Virtual Design concept in this paper is slightly different. Since building a Brake System model and performing Complex Eigenvalue solution is time consuming, it is still not efficient if we do not know what kinds of design changes are effective but just randomly modifies the model for design studies. Therefore, authors propose a method that uses the baseline model to perform sensitivity study and use the calculated sensitivity data to determine optimum combinations to improve squeals noise. The whole optimization process does not involve re-solving the model and can establish modal frequency tuning targets for brake structure. With the modal target, engineers can then focus on how to modify the structure FE model to achieve the modal tuning target to suppress Brake Squeal problem. Authors named the whole optimization process as Virtual Design Studies of Brake Squeal. The common Brake Squeal analysis process is shown in Figure 1. Before the model can be used for design studies, the system model needs to be correlated with test data. During system model correlation stage, many iterations are necessary so as the Complex Eigenvalue Solutions. After the system model is correlated, the current Complex Eigenvalues are used as a criterion to determine the brake system instability but does not provide how to change the structure to improve the system unstable modes. Engineers then modify the model with no clear design direction and re-run the Complex Eigenvalue solves until stable system Complex Eigenvalues are found. During this design study process, a lot of models need to be built and many Complex Eigenvalue Solutions are involved. Building a Brake System model and solve Complex Eigenvalues are very time consuming so it will take a long time to complete a whole development cycle.

VIRTUAL DESIGN STUDY CONCEPT

Brake FE Model

NASTRAN Complex Eigenvalue Solve

easily without resolving the system model. Then the sensitivity values are used to determine how much frequency shift of critical modes or mode coupling strength reduction to avoid unstable mode at operation coefficient of friction. Engineer can analyze the critical modes and determine effective modifications can reach their frequency target. Hence, the brake squeal design studies are no longer trial-and-error iterations but with specific design directions. Once the critical modes achieve target frequencies, Complex Eigenvalues will be compared to baseline results to verify its result.

Test Data

correlation
Bra ke FE Model

FE Model of Design Study

NASTRAN Norma l Mode Solution

NASTRAN Complex Eigenvalue Solve

MATLAB Eigenvalue Solution

Stable Complex Eigenvalues

Test Data

correlat ion

Figure 1 Common Brake Squeal Design Study Process.

Stable Co mple x Eigenvalues

Critica l mode pa ir identificat ion and sensitivity calculation

The proposed Virtual Design Study process is shown in Figure 2. Due to the new Modal Domain Analysis approach, Complex Eigenvalues is no longer obtained by using NASTRAN Complex Eigenvalue Solver. The Complex Eigenvalues can be calculated using MATLAB Matrix Eigenvalue Solution with the inputs of Normal mode frequencies and mode shapes of pad-rotor contact Degrees of Freedoms of non-friction Brake System model. The detail discussion of formulation will be discussed in the following section. The new Complex Eigenvalue evaluation process reduces significant solution time especially trying to establish detail Eigenvalue trace shown in Figure 3. The new method only needs to solve system normal modes once and use MATLAB to calculate all Eigenvalues. Comparing to the current approach which needs to run Complex Eigenvalue solve for every data point, the time saving for just generate a Complex Eigenvalue trace plot is tremendous. The new method uses the natural frequency trace such as Figure 3 to identify the critical modes that converge and create unstable system Complex Eigenvalues. Since the new analysis method can provide closed-form Eigenvalue solution, the sensitivity of critical mode trace to the change of coefficient of friction can be estimated

MATLAB Eigenvalue Solution

Determine Critica l Mode Targets frequency and coupling Strength

NASTRAN Norma l Mode Solution

Change FE Model to meet Norma l Mode Target Freq

Figure 2 Proposed Virtual Design Process

The brake system with no damping and friction can be expressed as

[M ]{u}+ [K ]{u}= {0}

(5)

The modal domain transformation can be obtained for equation (5),

{u}= [ ]{ }

(6)

where is the modal matrix of system equation (5). With the modal transformation, the following relationship (7) can be established.
12 0  0 2 0 2  T , [ ] [K ][ ] =   0 2 0  0 n

[ ]

Figure 3. Imaginary part of complex Eigenvalue of a Brake System.

[ ]T [M ][ ] = [I ]

(7)

where j is the j-th natural frequency of system

MODAL DOMAIN FORMULATION


The brake system equations of motion can be expressed as

equation (5). From (7) and (4), the Complex Eigenvalue of a non-damping Brake System with friction can be expressed as

[M ]{u}+ [C ]{u}+ [K ]{u}= {F f }

(1)

where, M , C , and K are the mass, viscous damping, and stiffness matrices respectively for the non-friction system, u is the displacement vector of the system, F f is the friction force applied to the system. A dot over the displacement denotes differentiation with respect to time. The non-friction system is defined as the brake system, which has its pad and rotor in contact but the friction force is zero. The friction force is modeled as

12 0 2 s 2 + 0 2   0 0

0  0 }= { 0} (8) + K s f {   2  n 

[ ]

The system Complex Eigenvalue is expressed as s and the dynamic friction term becomes

[ ] [
f

T friction force direction contact DOF

[K ][
f

normal direction contact DOF

(9) The inputs of system equation (8) are natural frequencies of non-friction system, the mode shape of non-friction system at contact degrees of freedoms and the value of coefficient of friction. Therefore, establishing root locus plot only needs to solve the system model once and then use numerical tool such as Matlab to calcuate complex eigenvalues with different values.

{F }= ({N }+ {N
f static

dynamic

})

(2)

where and N are the coefficient of friction and normal force respectively. For the Eigenvalue problem, the static force, which is due to the pressure applied by the piston, is removed from the equation of motion. The dynamic normal force is caused by the vibration of the pad and rotor and is represented by

VIRTUAL DESIGN APPLICATION


A brake system model is shown in Figure 4.

{N

dynamic

}= K ({u
s

N , rotor

} {u

N , pad

})

(3)

where

u N , pad denotes the displacement of pad in normal


Hence the system equation of motion

direction. becomes

[M ]{u}+ [C ]{u}+ [K ]{u}= K s [K f ]{ u}

(4)

From Figure 3, there are two critical modes at 3150 Hz and 3200 Hz converging and create an unstable Complex Eigenvalue at about 3200 Hz. The middle mode at 3175 Hz is close to two critical modes but it does not have strong interaction with these two critical modes and it remains at same frequency as increases. In order to fix a Brake Squeal problem whose coefficient of friction 0.5, as shown in Figure 5(b), we either reduce the convergence speed of two critical modes or further separate these two modes. However, the questions are how much mode separation is necessary or how much convergence speed reduction is necessary to fix the problem. MODE CONVERGENCE SPEED
Figure 4 Brake Squeal System model.

To quantify the critical mode convergence speed, we can take advantage of the symbolic form of the system equation of motion (8) and calculate the closed form solution of The first process of virtual design is system model analysis. The test data and the model Complex Eigenvalue correlation are shown in Figure 5.

convergence speed

2 dsij

d 2

( ) [ ] * [ ]
f ij f

ji

(10)

Note that the convergence speed is mainly controlled by the mode shapes of the system because f is made

[ ]

from the mode shapes of rotor-pad contact degrees of freedom. Virtual Design Approach It is time consuming to generate a Brake System model and it is also time consuming to make a design modification of component to re-construct a modified brake system model. In todays industry environment, the product development cycle is cut down significantly. Hence, the old trial-and-error method is no longer suitable. It also applies to the Brake Squeal problem solving, brake industry definitely needs a quick turnaround re-design method. The proposed Virtual Design method can definitely cut down the trial-and-error time and provides a clear design modification target for fixing Brake Squeal. The virtual design concept is based on equation (8) and (10). Equation (8) shows that the system Complex Eigenvalues are functions of non-friction natural frequency i and mode shapes , and equation (10)
Figure 5 The figure above shows the Brake Dyno test data and squeal around 3.2 KHz; the red dot represents a squeal event. The lower plot shows the root locus plot of the system and unstable Complex Eigenvalues at 3.3 kHz. Note that the straight line of the lower plot represents the system reference damping value.

shows that the convergence speed is controlled by the product of f ij * f ji . There are two ways to

[ ] [ ]

perform the virtual design study. The first virtual design is to evaluate the improvement of separation critical modes. For example, moving the critical mode of Figure 3 from 3150 Hz to 3100 Hz such that the mode convergence may happen at higher coefficient of friction to create a more stable system. The

way to perform this virtual design study is reassign the value of i = 3150 Hz to i = 3100 Hz of equation (8) and recalculate the system Complex Eigenvalues. The reason we can make this change is based on minor changes of the structure, such as removing some material, may shift natural frequency and make minor influence on its mode shape. Therefore, we can use Equation (8) to calculate the virtual design Complex Eigenvalues. Since all information need for Equation (8) are already existed, we do not need to re-solve the system model so it does not increase any calculation time. The virtual design result compared to the baseline result is shown in Figure 6. The critical separation by 1.5 % is an efficient way to suppress the system instability. To achieve the 1.5 % natural frequency separation target. We can examine the mode shape and strain energy of these two critical modes. The first critical mode of 3150 Hz is shown in Figure 7, and its energy is mainly at caliper. Engineer can examine the mode shape and the high strain energy area to determine the most effective changes to shift the natural frequency to lower frequency. The second critical mode is shown in Figure 8 and its strain energy is in rotor component. Hence, modifications on rotor may move the frequency to higher frequency.

Figure 7 The first critical mode.

Figure 8 The second critical mode

In a real brake structure, it is a highly integrated structure and only certain areas are allowed for minor changes. With critical mode natural frequency target, engineers can spend time to focus feasible and effective modifications for design modifications. The second way to perform the virtual design study is to reduce the mode convergence speed. From equation (8), we know that the convergence speed is proportion to the product of f ij * f ji . If we change the value of

Figure 6 The comparison between the baseline system and virtual design system. The red circle is the baseline system and blue diamond is the virtual design result.

[ ]
f

[ ] [ ]

ij

, we can change the convergence speed.

For

example, if we want to reduce the critical mode convergence speed of Figure 3 to one-half, we can set

[ ] = 0.5 * [ ]
f ij f

ij

(11)

The cross mode coupling mechanism is discussed in authors another paper 2003-01-1624. The reduction of mode convergence speed can be obtained by reducing the vibration level of rotor shell modes or reducing the mode shape similarity between coupling modes. Case 2 shows brake system has squeal problem at 5 kHz squeal problem as shown in Figure 9, and the highest mode convergence happens at mode 67 and 69.

Using the virtual design approach, we found that increasing mode separation by 2 % is not effective for this squeal problem as shown in Figure 10, because this problem is mainly due to strong mode coupling. Hence, it is only effective by reducing the convergence. To reduce the convergence speed by 50%, we set

[ ]
f

67 , 69

= 0.5 * f

[ ]

67 , 69

in equation (8).

(12)

The modified system is shown in Figure 11 and the unstable mode is brought into more stable region. The critical modes of the structure is mainly pad and rotor modes as shown in Figure 12a,b. One is rotor butterfly mode and rotor bending mode. The design modifications should then focus on how to reduce their vibration level to fix squeal noise.
Figure 11 The virtual design result by reducing coupling strength.

Figure 9. A brake system with 5 kHz squeal showing in Root locus plot.

Figure 12a (upper plot) shows the pad critical mode; 12b (lower plot) shows the rotor bending critical mode.

CONCLUSION
The paper presents a new design study process for brake squeal development. The Computation efficiency of the new design process is compared with current commonly used procedure. Based on the modal domain formulation, the Complex Eigenvalue characteristic equation can be expressed in symbolic form so the mode converging speed can be derived in closed-form solution. Since the System Complex Eigenvalues are functions of natural frequency and friction coupling matrix of the nonfriction system, we can make virtual design changes of critical natural frequency and friction coupling matrix to evaluate the improvement of virtual designs. After the

Figure 10 The virtual design result compared to baseline result.

satisfied improvement is obtained in the virtual design, the structure design modification target can be determined, engineers can focus on effective design changes to achieve the squeal suppressing goal.

REFERENCES
1. Chung, Chih-Hung, and Steed, William, and Kobayashi, and Nakata, Hiroyuki, A New Analysis Method For Brake Squeal Part I: Theory for Modal Domain Formulation And Stability Analysis, SAE paper 2001-01-1600. 2. Nakata, H., Kobayashi, K, Kajita, M and Chung C., A New Analysis Approach For Motorcycle Brake Squeal Noise and its Adaptation, SAE paper SETC2001-01-1850. 3. Liles, G., Analysis of Disc Brake Squeal Using Finite Element Methods SAE paper 891150 4. Chen, L.W. and Ku, D, Stability of Nonconservatively Elastic Systems Using Eigenvalue Sensitivity, ASME Journal of Vibration and Acoustics, Vol. 116, No. 2. 5. Ghesquiere, H., Brake Squeal Noise Analysis And Prediction, ImechE paper No. 925060. 6. Lewis, T., Analysis and Control of Brake Noise, SAE Paper 872240. 7. Chung, Chih-Hung, and Donley Mark, Mode Coupling Phenomenon of Brake Squeal Dynamics, SAE Paper 03-NVC-386.

S-ar putea să vă placă și