Fillet Fails - Complex curve

[tfarllc]

The fillet function fails.  I created an extrusion using lines from a dxf file.  I closed the curve first then created the extrusion (see Thicker padding extrusion in the attached model).  Next I tried to create a fillet (1.9 mm) using the command fillet command in the 3D Model tab.  I probably got the fillet to go 80% of the way around.  Where the fillet failed I would try smoothing the sketch curve that creates the extrusion.  I actually made the problem worse.  

   Any ideas on how to create fillet all the way around  the edge of "Thicker padding" extrusion?

I tried sweeping but that also failed.

Using - AutoDesk Inventor 2017

[Anonymous · Accepted solution]

With such a complex design, I would first look at the DXF data. Is it indeed a closed loop. Any chance you have multiple lines that cross back on themselves. Things I have done is try to break up the sketch into sections and perform what you want to do. If it works, then undo back and try a different section to find where it breaks. The fact that the top fillet is in a section tells me that your profile is not closed or well defined. I don't mean dimensioned, just not well defined.

[tfarllc]

Thank you for your input.   The dxf was a mess to start.  I picked "close loop" and closed it.  Additionally, I tried fillet and create arc's  inside the sketch mode to smooth the curve.   This method gave me hope but what gains I made I started losing.

     In that model I posted there is Extrusion 5 - Spline Curve.   I hand sketched a spline around the dxf curve.  The spline looks smooth except for the start/end point.  I could not fillet the Extrusion 5 either  so I suppressed it. 

   I am a bit confused.  I can't seem to pick a small enough section of the curve to fillet.

[tfarllc]

The best solution at this point was to use the hand sketched spline curve and extrude.  This curve left an irregularity at the start/end point.  I put some effort into smoothing the spline there.  Next,  I created a fillet with a tiny radius (.1 mm).  Slowly I increased the radius.  Where the radius failed I would reduce the increment and try again.  This method resulted in something of practical use but a mine field to someone trying to alter the model.