This presentation was made at the 2019 NAFEMS World Congress in Quebec Canada

Resource Abstract

The accurate conversion of time-based data into an equivalent frequency domain form has become very important because of the increased use of frequency domain analysis for random response and durability calculations. Typically, for road vehicles, the loading conditions are measured at the wheel center and then cascaded into body loads using multi body dynamics systems (a partially virtual model). Or, in a more sophisticated system, the road profiles could be combined with a wheel/tire model to transfer the loads through the wheel system and onto the body of the vehicle (a fully virtual model). In both simulations, the cascaded loads on the body are in the time domain and usually in the form of multiple channel time signals. The correlation between the multiple channels can be very important to the final outputs for random response and fatigue.



If a frequency domain analysis is required, this time data must be converted into the frequency domain. This involves well understood “Fourier” conversion techniques in the form of Fourier Series or Fourier Transforms. However, a typical analyst faces 3 significant barriers when performing this conversion process.



Firstly, the frequency domain approach itself requires adherence to limiting assumptions. The data being processed must be,

• Stationary

• Gaussian

• Random



Typical users struggle to quantify these assumptions.



Secondly, the legacy use of Fourier Transforms (such as the Fast Fourier Transform – FFT) is difficult for a typical User because of several variables which must be set (such as FFT window shape, FFT window length, etc.). Setting these (and other variables) requires prior experience, which is beyond a typical User. Furthermore, the variables usually have to be set one by one for each load set (event) and this can be impractical.



And thirdly, the correlation between numerous (sometimes 100’s) of channels, and the channel mapping to the model, must be retained which represents a significant data management and usability issue.



Recent technical breakthroughs have solved most of these issues, making the process of loads conversion possible for a typical User. The first part of the solution involves a sophisticated process for “Loads Conditioning” prior to the Fourier Transformation. The second part involves an automatic choice for the FFT window length. And lastly, the mapping between load channels, their correlation, and the FEA model subcases must be properly handled. Taken together, these 3 improvements represent a significant breakthrough related to the task of loads conversion from time domain to frequency domain. This paper highlights the improvements for many different forms of loading data.