Simcenter Testing Solutions How to perform a Coherence Analysis for Road Noise?
2021-06-01T09:03:39.000-0400
Simcenter Testlab
Summary
This article will explain how to perform the coherence analysis in the context of Road Noise, to separate the sources - front x rear / solid x airborne.
Details
Introduction:
Usually, a Road noise analysis starts with a processing called Coherence Analysis with the objective to identify if the noises perceived by the driver are mainly generated via Airborne or Structure-borne paths or even one might need to understand if a noise peak in a certain frequency is coming from the Front or Rear suspension.
The Crosspower calculation in Testlab will filter the data based on the references selected. The results will be only what is coherent with the references. After the calculation of the Crosspower, one can apply Principal Component Analysis in order to calculate the Coherent Autopowers, which are the part of the Autopower, of a reference or response, that is coherent with the virtual references (uncorrelated signals).
The process for Coherence Analysis is as follow:
Example:
Using the Road Noise analysis as an example, common instrumentation is:
Using the Road Noise analysis as an example, common instrumentation is:
- 2 Internal Microphones à 2 channels à Responses;
- 1 Triaxial Accelerometer per Knuckle: 12 channels à References;
- 2 Microphones near to each tire: 8 channels à References;
With this instrumentation, constant speed measurements were made for 25 seconds.
For the Coherence Analysis processing, load in Simcenter Testlab Desktop the Add-ins: Signature Throughput Processing, Multi-Reference Post Processing and Principal Component Analysis.
First, add the Run into the Input Basket and replace it in the Time Data Selection workbook. On the Time Data Processing, select all channels for processing.
Select the acquisition parameters as follow:
First, add the Run into the Input Basket and replace it in the Time Data Selection workbook. On the Time Data Processing, select all channels for processing.
Select the acquisition parameters as follow:
It is important to use the entire time trace to have the maximum number of averages. In this case, the overlap also helps. Choose the Frequency Resolution as you needed it.
Keep in mind that for Multi-Refence processing it is necessary that for the different groups (Acoustic, Vibration, Other) you will need the same Sampling Frequency. If in your case, Acoustic and Vibration channels have different Sampling Frequency you might need to resample the signals.
Set the Channel Processing as follow:
Keep in mind that for Multi-Refence processing it is necessary that for the different groups (Acoustic, Vibration, Other) you will need the same Sampling Frequency. If in your case, Acoustic and Vibration channels have different Sampling Frequency you might need to resample the signals.
Set the Channel Processing as follow:
As function select Crosspower Power and select all the channels as references apart from the two microphones inside the cabin. The same should be applied to the Acoustic and Vibration groups.
Once this is done, press calculate.
After the calculation, a new Run will be available. Change to the worksheet Multi. Ref. Processing and select the new run as data source.
Once this is done, press calculate.
After the calculation, a new Run will be available. Change to the worksheet Multi. Ref. Processing and select the new run as data source.
Change the Data Set visualization to Matrix – Per Run.
For the first calculation, we will use all the references calculated. In this way, move to the Calculate minor worksheet. On the PCA (Principal Component Analysis) tab, select the Referenced Virtual spectra and the Coherent Autopower. Give it a Postscript to identify the new run and press Calculate.
Come back to the Data Selection minor worksheet and use the Demote DOF and Promote DOF to keep only the 8 external microphones as References. Move to the Calculate tab, give it a new postscript and press the calculate button.
Repeat this process, but selecting the references below:
- 12 Knuckle vibration channels;
- 6 Front Knuckle vibration channels;
- 6 Rear Knuckle vibration channels;
Once this is done you should have a total of 5 Runs with the PCA’s calculated.
Analysis:
Airborne x Structure Borne
On the Navigator, create a FrontBack picture, or the one that you prefer, to plot the results. The first analysis will be an evaluation of the Airborne x Structure-borne phenomena.
Let’s first plot the measured Autopower for our target at microphone FRLE:01, as a result of our processing on Time Data Processing. Then, plot for the same microphone, the resultant Coherent Autopower curve that used all the 20 references. The differences between these two curves are due to references or phenomena’s that were not considered in the measurements, like the engine noise, or even aeroacoustics.
Then, inside the PCA run calculated with only the 12 knuckle vibration channels are references, select the Coherent Autopower for the microphone FRLE:01 and plot it. Plot the same Coherent Autopower, but for the Run that used only the 8 external microphones and plot it in the same picture. In this way, we can evaluate, for each frequency peak or frequency range, if the noise has a Structure or Airborne path.
Airborne x Structure Borne
On the Navigator, create a FrontBack picture, or the one that you prefer, to plot the results. The first analysis will be an evaluation of the Airborne x Structure-borne phenomena.
Let’s first plot the measured Autopower for our target at microphone FRLE:01, as a result of our processing on Time Data Processing. Then, plot for the same microphone, the resultant Coherent Autopower curve that used all the 20 references. The differences between these two curves are due to references or phenomena’s that were not considered in the measurements, like the engine noise, or even aeroacoustics.
Then, inside the PCA run calculated with only the 12 knuckle vibration channels are references, select the Coherent Autopower for the microphone FRLE:01 and plot it. Plot the same Coherent Autopower, but for the Run that used only the 8 external microphones and plot it in the same picture. In this way, we can evaluate, for each frequency peak or frequency range, if the noise has a Structure or Airborne path.
Front x Rear
Narrow our analysis to only the Structure-borne phenomena, we can evaluate the influence of each part of the suspension, Front versus Rear. For this case, after plot the same 2 first curves, as the previous analysis, we can plot the internal microphone Coherent Autopowers for the Runs with only the front vibrations as references (6 references) and the rear vibration as references (6 references).
Thus, we can compare which part of the suspension, front or rear, drives the peak of noise inside the cabin, per frequency.
Narrow our analysis to only the Structure-borne phenomena, we can evaluate the influence of each part of the suspension, Front versus Rear. For this case, after plot the same 2 first curves, as the previous analysis, we can plot the internal microphone Coherent Autopowers for the Runs with only the front vibrations as references (6 references) and the rear vibration as references (6 references).
Thus, we can compare which part of the suspension, front or rear, drives the peak of noise inside the cabin, per frequency.
Conclusions
With this simple instrumentation and operational data, it is possible to extract relevant information about the road noise phenomena, providing insights about the frequency range of the Structure-borne and Airborne, and front and rear suspension.
This approach is not only applicable for road noise but anywhere where you might want to separate partially correlated or fully uncorrelated sources.
With this simple instrumentation and operational data, it is possible to extract relevant information about the road noise phenomena, providing insights about the frequency range of the Structure-borne and Airborne, and front and rear suspension.
This approach is not only applicable for road noise but anywhere where you might want to separate partially correlated or fully uncorrelated sources.
Associated Components
Simcenter Testlab Digital Image Correlation
Testlab Environmental
Testlab Acoustics
Testlab Data Management
Testlab Desktop
Testlab Durability
Testlab General Acquisition
Testlab General Processing & Reporting
Testlab Rotating Machinery & Engine
Testlab Sound Designer
Testlab Structural Dynamics
Testlab Turbine