Automobile acoustic modeling based on AUTOSEA simulation software

For many automakers, the complete acoustic modeling of a vehicle is still a dream. However, acoustic simulation methods are becoming more and more widely used and are becoming an important design tool in the effort to reduce development time.

Acoustic modeling is often misunderstood as a magic tool that can solve all problems. In fact, to date, acoustic and vibration modeling can only provide important suggestions rather than definitive answers, and it must be considered as a problem-solving tool during the development and prototyping phases.

Because people don't believe in it, not all the acoustic problems of the vehicle are taken into account during the design phase, which leads to acoustic problems appearing at the prototype stage or later. If the experimenter has access to the FEM (finite element method) model, the acoustic problems can be considered from the beginning; and if the designer can understand the true meaning of a measurement report, the problem is easier to solve. Therefore, acoustic modeling should be an additional tool combined with prototype development and problem-solving oriented, and the relevant process can follow the following principle steps.

During the design phase: 1. Obtain a simplified acoustic FEM model; 2. Evaluate the noise level using the BEM or SEA method under estimated input forces; 3. Calculate whether there will be serious problems during the design phase.

During the prototype phase: 1. Obtain experimental data and isolate noise issues from the prototype; 2. Obtain a cyclic model for each problem and check the input force amplitude; 3. Try possible solutions and simulate the expected results; 4. Verify the solution applied to the prototype; 5. Use experimental data to improve the solution.

Design Methodology

The following is a case cited from the Vibro-Acoustics Science Inc. Application Note, a journal on vibro-acoustics, which describes the application of AUTOSEA simulation software to vehicle interior noise (see Figure 1).

Figure 1 AUTOSEA model of a vehicle with subsystems

Typical questions for interior noise are: the level of interior noise; the sources and transfer paths of noise; how to reduce the noise level. In order to answer these questions, it is necessary to import a FEM model of the vehicle. This is a typical coarse mesh model of the "concept phase", which generates about 150,000 elements, which of course need to be reduced to about 50,000. The simplified model must be revised to solve certain problems and then check whether the original characteristics are still maintained.

Now it is possible to evaluate the noise level at multiple receiving locations in the vehicle using a simplified model using a vibroacoustic approach (based on the frequency range of interest). Two methods are available: FEM-BEM (finite element and boundary element) up to 200 Hz and SEA (statistical energy analysis) from 200 Hz and above. FEM-BEM can be applied to both structure-borne and airborne paths, but it becomes complex and requires too many elements when the model density rises rapidly, while SEA gives good results when the model density increases.

To identify the vibration or noise source and transmission path, the relevant technology must be used in conjunction with the FEM-BEM method, so that noise reduction technology may be able to produce results. The SEA method is based on energy transfer calculations to identify the contribution of each source and the efficiency of the sound wave transmission path (see Figure 2).

Figure 2 SEA network

Complex problems can be solved more efficiently by analyzing them one by one. Here we show the use of AUTOSEA in the specific case of considering the roof noise (generated by aerodynamic pressure).

First, we imported the FEM model and stored the material and beam section properties in the database, then simplified the geometry complexity by creating only the NASTRAN roof element and all SEA substructures (except for the windshield which is visible). We added curved plates to the SEA model to represent the roof.

1. Combine the roof, windshield and interior cavity and observe the number of modes in the band, as the acoustic cavity increases very quickly.

2. Plot the wave number of the selected subsystem and observe that the coincidence frequency of the glass is much lower than that of the roof. In a wide frequency band, the windshield is an obvious radiator.

Connecting the roof and windshield subsystems, the angle is close to normal, so energy can only be transferred through moments. Connecting the roof and interior subsystems, the radiation efficiency of the roof panel depends on the fixed boundaries (such as the general baffle and roof lining respectively increase and decrease the radiation efficiency). The effect of the roof lining on the radiation efficiency can be analyzed and simulated by the SEA method or by importing experimental data into the model.

[page]

A power source is connected to the canopy subsystem to replace the wind (aerodynamic) pressure source. The spectrum of the aerodynamic pressure can be determined from wind tunnel or road test data, or calculated as a default spectrum. The power input to the canopy is because wind noise is large at low frequencies and decreases rapidly due to the divergence between the structural and convective wavenumbers. Note that the other vibration noise sources in the model (see Figure 3) are the structure-borne engine noise (represented by vibrations measured at the vehicle front rail) and the airborne tire noise (represented by the diffuse sound pressure level measured at the bottom surface of the vehicle floor).

Figure 3 Vibration noise source

If you want to solve the above network to get predictable A-weighted sound pressure values ​​for cars, trucks, underbody, windshield, and headliner, then you might notice that the higher measurement levels of the A-weighted sensor are for frequencies above 500 Hz. If you look at the energy input to the interior of the vehicle (for the multiple vibration noise source problem), you will find that tire noise occupies the frequency range of 250 to 1,000 Hz, but at higher frequencies it is the windshield that is responsible for the interior noise. So now we have to understand what is causing the windshield to vibrate.

The first approach was to perform “source ranking” by freezing the interior sound pressure, then disconnecting the wind and tire noise sources and resolving all the issues. The results showed that the high-frequency radiation from the windshield into the vehicle was caused by structural excitation caused by engine vibration.

We reconnect the wind and tire noise sources (the inputs to the floor mass method are 0.73m2 and 4.5kg) and re-solve the problem to obtain the energy flow at 2000Hz. As expected, the windshield and instrument panel are the two main contributors to the interior noise, but now we also have information about the structural path through the upper and lower pillars.

The transmission of tire noise in the frequency range of 500 to 2 000 Hz can be reduced by adding a floor mat. However, to characterize the floor mat, a 3-layer sample must be created in the database and added to the connector between the tire noise and the interior of the vehicle. The composition of this component can be tested and adjusted to achieve the maximum effect (within the desired frequency band). The double wall resonance effect may increase the interior sound pressure in certain specific frequency bands, such as 250 Hz. For this model, other design solutions can also be tested, such as: optimization of the floor mat; optimization of the interior sound absorption kit (also using the SEA method); changes to the structure of the front section of the vehicle to reduce high-frequency noise caused by engine vibration.

Prototype Approximation Method

Once a prototype or parts of a prototype become available, some preliminary investigations can be carried out and after obtaining representative data and a frequency analysis (a “listening” session of the interior noise), standard measurement procedures can be carried out to combine the technical understanding of the data with the physiological perception of what is going on.

"Listening" is the key point, and this feature can actually be added to measurements made with an artificial head or a BHM (Binaural Measurement Microphone from HEAD Acoustics).

By paying attention to the noise activity in time-frequency representation, the acoustic engineer can focus on some specific noise he notices and use HEAD Acoustic's SQ-LAB to enhance the noise or suppress other spurious phenomena. For example, we can listen to the original recording and the filtered version of the same recording at the same time; filtering means reducing the amplitude of a specific component of the sound and feeling the change in the sound. This kind of sensory analysis is very useful for establishing cause and effect, identifying single noise sources, and identifying possible transmission paths of structure & media.

The vehicle is usually tested in both accelerated and stationary states. The first step is usually to identify problems on the test track, such as the maximum A-level in the car, the key areas of roar, noise components, rotation speed and speed, and sound quality. Listening and analysis of the road test will provide prompts for the definition of the subsequent "test plant", which includes the identified problems, priorities and investigation steps.

The acoustic model must be investigated first, so that the validity of the investigation can be verified more quickly and the experimental validation can be carried out with more confidence. Perhaps manual modification of the acoustic model does not always give the desired results, because simulation tools are not perfect, but in any case, some suggestions can be obtained to guide the experiments. It is this parallel investigation of the acoustic model and the prototype (a bilateral exchange of information) that accelerates the exploration of possible solutions and their validation for a specific noise problem.

Modal Analysis

To outline the guiding significance of acoustic models for experimental testing, we illustrate some representative tasks performed in a real case of a light truck.

Experimental modal analysis of the entire vehicle is not an easy task and does not provide a clear check on the FEM model. The problem is that the excitation of the suspension and structure alternately shows high damping for the high modes and nonlinearity for the low modes. It is more efficient to perform modal analysis on the substructure (windows, roof, doors, etc.) since the acoustic model report shows the type of problem that needs to be investigated. Try to keep the excitation points close to the real situation, for example: engine or cabin suspension, roof distributed forces, etc.

Acoustic Mode

The experimental determination of the acoustic modes in a car interior has two purposes: first, to check the theoretical modes in order to improve the acoustic model, and second, to measure damping (the experimentally determined damping values ​​are attributed to the acoustic material, the dashboard and the seats). The problem here is the acoustic excitation of the interior cavity, since some of the types of loudspeakers used do not support receiving any information about the exciting forces, such as the system input. One possibility is to use a standing wave tube, which is often used to determine the acoustic properties of materials. One end of the standing wave tube is mounted on the car window (open), and the other end is equipped with a loudspeaker. Two microphones are mounted in the longitudinal direction. By measuring the transfer function between the two microphones, the energy flowing into the cabin from the other end of the tube can be obtained.

Input force

Road tests are used to measure both noise and vibration; while noise measurements are mostly concerned with the noise level inside the vehicle, vibration measurements of the engine mounts provide the amplitude and spectrum of the input forces for structure-borne noise. By inserting the measured forces into the acoustic model, it is possible to more accurately conclude that the noise calculation is based on the input vibration.

As for airborne input forces, such as the acoustic radiation of the engine, a different approach (depending on whether a FEM model of the engine is available or not) is necessary. In most cases, this cannot be known at the design or prototype stage, so it is interesting to determine the acoustic model of the engine experimentally. The process is in fact quite self-explanatory: the mapping of the hypothetical sound pressure level SPL of a hypothetical surface around the tested engine allows to define the distribution of the sound pressure and the dependence of the phase on the frequency. When given the same distribution, an equivalent acoustic model of the source is obtained by lagged least squares estimation.

An acoustic model of an engine requires 10-20 elementary pistons with radiating areas (weighted engine geometry). The process is easy to understand, but requires good skills in execution: first, a multi-channel measurement system must be set up (with enough microphones to obtain phase relationships and some reference microphone positions); second, a critical speed must be considered as a static condition in the measurement, and this information can only be obtained through analysis of road tests and "listening" sessions; third, this information can be inserted into the vehicle's acoustic model as a real air-borne input force, and the interior noise level can be recalculated.

Conclusion

The study of vehicle acoustics is a team effort, not the application of individual feelings and expert opinions. There are several possible approaches, both at the design and prototype stages, and given the many possible noise sources, there is no single exact procedure that works for every subject.

If you want to take full advantage of modern computing systems, you should follow some simple methods to reduce the complex overall project into independent problem groups. Don't get lost in thinking about the vehicle acoustics problem as a general problem, as the old saying goes "don't miss the forest for the trees".

For example, from the beginning of the design phase, simplified methods should be introduced into the acoustic study of the vehicle, and when the prototype is available, the acoustic model must be adjusted in combination with experimental data. Once the noise problem is identified experimentally, the possible solutions should be run many times on the acoustic model adjusted in combination with experimental data, because it is faster and cheaper to make changes and verify them on the prototype than to make changes and verify them on the prototype.

Optimal results and rational development of the vehicle from an acoustic point of view are based on this exchange of data between experimenters and modellers.