Approximate Prediction of Multimodal Crest Statistics and System Reliability for Impulsive Noise Loading

摘要

The current analytical models of impulsive noise and system response thereto are found numerically intractable, indicating a need for engineering approximations. One, based on finding an 'equivalent' unimodal system having approximately the same response statistics--here the crest statistics--is examined. It is found that fair agreement obtains, making it possible to use two reasonably accessible parameters to characterize any linear multimodal system response and thus predict its crest statistics systematically, provided that 'standard' curves of unimodal response crest statistics are available for statistically identical forcing. Limitations of this promising approach are explored, encouraging its use in reliability prediction pending further studies. Also, the need for further studies of impulsive noise response crest distribution 'tails' is noted for prediction of overload and wearout reliability on the basis of wear-dependent failure rate concepts, since extrapolation of empirical data into the 'tail' (here done on the basis of a Weibull crest fit) is increasingly necessary the more 'impulsive' the response is (i.e., the more the parameter is less than unity). (Author)