I recommend you DON'T put fillet radii on the corners of square sectioned springs. This will seriously slow down your model for very little benefit. This is particularly important if you use a lot of springs in your assemblies. I put my springs on their own layer
If you want to get flash you can link all the pitches together via relations so as the spring compresses, the coils get closer together and if you build in the spring constant you can get Pro/E to calculate the loads at any particular compression
Sorry but I am not prepared to publish the spring model above as it stands. What you can't see is that it is derived from a family table that is 3 layers deep and currently contains over 500 instances.
I have far too much time invested in it to simply give it away.
I haven't downloaded your file, but I think the problem you're facing is pretty generic and one I have faced in the past.
I think you are trying to achieve 'closed and ground ends'. You need to use the variable pitch option on your helix to close the ends. This is a little tricky in Pro/E, because you cannot have the colis overlap as you reduce the helix angle (pitch): you cannot have the pitch less than the wire diameter. The spring pitch curve is trapezoidal in shape in this example: while Pro/E will do some smoothing at the transitions in the pitch, it still isn't perfectly smooth, and you will have to live with this. Once you have achieved the closed ends, you need to create an offset plane at the ground end of each end of the spring, and make a cut there to finish the spring.
When I last did this in Pro/E about seven years ago, the variable pitchoption on thehelix generation feature was not very robust, and it took a number of iterations before the spring would be successfully modeled without regeneration failures. Hopefully the stability of this function has been improved since then.
Variable pitch works fine, at least in the last 3-4 years I played with it.
Flatten springs by cutting away top and bottom is - in the field - only done for large springs because it's an extra fabrication step that takes money. Small springs are specified to have last part of the coil flat wounded. Variable pitch helix does not end flat because you can't have zero pitch. If you really want the last end flat you need to "knit" a revolved protrusion to both ends. For assembly sake it is easier to create a datum that is normal to the axis and tangent to the end of the spring (you need a datum curve at the end to attach to).
'Flatten' springs in the field, esp. large springs?? I cannot imagine anyone doing this. Most compression springs (large and small) have 'closed ends' because it increases the bearing area at the end of the spring and makesthe areamore uniform, resulting in a better load distribution. While it is true that you cannot have a zero pitch helix, the pitch of the helix can be reduced via the variable pitch option toa minimum. This is where the instability in the feature arises, as Pro/E is picky about just how close to zero pitch you can get. And just how close you can get depends on the geometry of the specific spring, thus the need to putz with it a bit in each example.
I do not mean to suggest this is a fatal flaw in the modeling feature, just a characteristic of it. This is probably due to roundoff errors that get ugly because the pitch is used in the divisor of some equation, thus resulting in a 'divide by zero' problem in the algorithm. The feature probably works a lot better in 64-bit systems than 32-bit systems. Every CAD system I have dealt with struggles with the pitch curve a bit when you try to close the endsof a compressionspring.
To trim the end of the spring (grind it flat), you need a datum plane normal to the helix axis and through a (datum) point at the end of the helix curve: it will not be tangent to the end of the helix curve.