Experimental and analytical investigation of further developments of Fatigue Damage Spectrum (FDS)

The Fatigue Damage Spectrum (FDS) is a popular method used in the industry to execute accelerated vibration testing for mechanical components and structures. This method uses acceleration test signals in time domain to derive vibrational velocity which in turn is used for obtaining induced mechanical stresses. Taking the SN-curve properties (slope and intercept) of the material and linear damage accumulation model (Palmgren/Miner) in account, the damage for the component is derived in frequency domain. The core of the process now comes into action by reducing the time of the test signal and preserving the damage content in each frequency band constant. The accelerated signal is converted back into time signal from frequency domain using a distribution function. This process ensures keeping the damage content in each frequency band constant while accelerating testing times on testbenches.
The process uses the relationship between vibrational velocity and mechanical stress to deduce the damage. Other vibrational parameters like acceleration for the dependency of mechanical stresses has also been investigated in recent times. However, the choice of parameter lies completely with the user. This study aims to aid the user in the choice of parameter by conducting experiments on an electrodynamic shaker and analysing the dependency of mechanical stresses on various vibrational parameters. As mentioned before, the slope of the SN-curve plays an integral part in determining the damages incurred. Here the question arises how would the results of FDS change if SN-curve parameters are varied for the same material (for example, parameters from FKM Guidelines, MIL standard or even from an experimentally determined SN-curve). The limits of FDS are investigated in this scenario. The transformation of the accelerated signal from frequency to time domain is undertaken with the help of a distribution function (often assumed to be Gauss) and a random phase distribution of the load amplitudes. In reality, loads are more often than not, non-Gaussian. From a new perspective, consideration of distribution functions like Lalanne, Dirlik as well as higher statistical moments like skewness and kurtosis are proposed for the reconstruction of the accelerated time signal. The research question arises here to which extent is the general assumption of a distribution function is valid and if necessary which additional information is required to achieve a better consensus between calculated and experimental results.
Experiments are conducted on an electrodynamic shaker with samples of structural steel and electro-grade copper. Stresses are measured using strain gauges and accelerations are measured using accelerometers as well as laser vibrometer. Test acceleration is done with the help of a data analysis software nCode from HBK. In parallel, FEM simulations as well as spectral methods of damage calculation are used to compare experimental results.