Fatigue assessment of vibrating rail vehicle bogie components under non-Gaussian random excitations using power spectral densities
Fatigue assessment of vibrating rail vehicle bogie components under non-Gaussian random excitations using power spectral densities
复制标题
DOI：
10.1016/j.jsv.2013.06.012
复制
发表时间：
2013-10
影响因子：
4.7
通讯作者：
P. Wolfsteiner;W. Breuer
中科院分区：
文献类型：
--
作者：
P. Wolfsteiner;W. Breuer
随机振动下的疲劳载荷评估通常基于载荷谱。通常，它们是基于时域信号用计数方法（例如，雨流）计算的。可替代地，方法是可用的（例如Dirlik），使得能够直接从对应的时间信号的功率谱密度（PSD）估计载荷谱;然后时间信号的知识是不必要的。这些基于PSD的方法具有巨大的优点，例如，如果信号评估来自基于有限元方法的振动分析的结果，则频域中PSD的仿真的计算时间远远超过时域中时间信号的仿真。这对于时域中具有非常长信号的随机振动尤其如此。基于PSD的振动模拟和基于PSD的载荷谱估计的缺点是它们对高斯分布时间信号的限制。与该高斯分布的偏差导致估计的载荷谱中的相关偏差。在这些情况下，通常只有计算时间密集的时域计算才能产生准确的结果。本文提出了一种处理具有真实的统计特性的非高斯信号的方法，该方法仍然能够使用具有计算时间优势的高效PSD方法。本质上，它是基于高斯分布部分的非高斯信号的分解。这些重新排列的信号的PSD然后被用于执行通常的PSD分析。特别是，详细的方法描述的时间信号的分解和PSD和交叉功率谱密度（CPSD）的推导，从多个真实的测量，而不使用不准确的标准程序。此外，其基本意图是设计一种通用的综合方法，该方法不仅能够在小的时间间隔内分析某个单一的载荷情况，而且能够生成代表性的PSD和CPSD谱，以取代时域中的大量测量载荷，而不会丢失疲劳载荷结果的必要精度。这些长测量值甚至可以代表铁路车辆的整个应用范围。所提出的工作演示了这种方法的应用程序受到随机振动引起的轮轨接触的铁路车辆部件。广泛的测量轴箱加速度已被用来验证这类铁路车辆应用程序的建议程序。
The assessment of fatigue load under random vibrations is usually based on load spectra. Typically they are computed with counting methods (e.g. Rainflow) based on a time domain signal. Alternatively methods are available (e.g. Dirlik) enabling the estimation of load spectra directly from power spectral densities (PSDs) of the corresponding time signals; the knowledge of the time signal is then not necessary. These PSD based methods have the enormous advantage that if for example the signal to assess results from a finite element method based vibration analysis, the computation time of the simulation of PSDs in the frequency domain outmatches by far the simulation of time signals in the time domain. This is especially true for random vibrations with very long signals in the time domain. The disadvantage of the PSD based simulation of vibrations and also the PSD based load spectra estimation is their limitation to Gaussian distributed time signals. Deviations from this Gaussian distribution cause relevant deviations in the estimated load spectra. In these cases usually only computation time intensive time domain calculations produce accurate results. This paper presents a method dealing with non-Gaussian signals with real statistical properties that is still able to use the efficient PSD approach with its computation time advantages. Essentially it is based on a decomposition of the non-Gaussian signal in Gaussian distributed parts. The PSDs of these rearranged signals are then used to perform usual PSD analyses. In particular, detailed methods are described for the decomposition of time signals and the derivation of PSDs and cross power spectral densities (CPSDs) from multiple real measurements without using inaccurate standard procedures. Furthermore the basic intention is to design a general and integrated method that is not just able to analyse a certain single load case for a small time interval, but to generate representative PSD and CPSD spectra replacing extensive measured loads in time domain without losing the necessary accuracy for the fatigue load results. These long measurements may even represent the whole application range of the railway vehicle. The presented work demonstrates the application of this method to railway vehicle components subjected to random vibrations caused by the wheel rail contact. Extensive measurements of axle box accelerations have been used to verify the proposed procedure for this class of railway vehicle applications.