Variance and Confidence Intervals for SEA Predictions

Statistical energy analysis has become an accepted method to predict noise, vibration and harshness in motor vehicles. SEA provides a statistical estimate of the vehicle response. In most cases, the mean response prediction is used to evaluate different designs. It has generally been found that SEA predictions of the change in mean acoustic response due to a specific design change are in good agreement with the change in the actual measured results. However, there may be significant differences between the absolute value of the predicted mean and the value observed from measurement. These differences are associated with the statistical variability of the prediction. This paper explores the use of the variance of the SEA prediction. Determination of confidence intervals for absolute prediction accuracy is described. In addition, the variance associated with SEA prediction of design changes is discussed. Methods to reduce the variance and improve the SEA prediction accuracy are presented.