What is the correct way to run a random response analysis when the current model has surface to surface contact in it? The results in this model are orders of magnitude high and there appear to be discontinuities at the transition from/to the edge of the contact area. I ran it as glued. The attached shows three layers of sheet metal in contact with each other. The bolts are spider rbe's, but are not visible in this view. The graphics don't show the elements on the curve part of the sheet metal. There are 5 elements around/through the bend.