The Surface Modeling Project







An account of Jeremy Singley’s efforts, with considerable technical assistance from Paul Johnson, to determine the limits of Autodesk Inventor as an industrial design tool. 












For now this site will consist of general notes describing what we’ve learned—sort of a diary to organize my thoughts.  Hopefully the site will later include videos showing actual screen moves demonstrating specific techniques, along with part files that the viewer can deconstruct.


Probably the best place to begin a discussion of surface modeling is with splines.  Splines can be thought of as elastic wires under tension, pinned to the sketch plane by coincident points.  If one segment of the spline is moved, its other segments will squirm in response down its length, but the pinned points will remain fixed.  This makes splines both useful and tricky to work with.


The spline user’s first rule of thumb might be:  Use the minimum number of points.  Common sense says it takes at least three points to describe a curve, but not so with splines.  Many of my splines have but two points—the beginning and the end.  Bow ties make this possible.  The “flat” bow tie can help make a clean transition from a straight line or a level surface.  The “curvature” bow tie is good for tweaking the depth of a spline’s wiggles on either side of a bow-tied point.  The “handle” bow tie allows the user to twist a spline point without changing the local radius or tension.  “Vertical,” “horizontal,” and “tangent” constraints applied to bow ties can make for smooth segues from coincident lines or curves.  Stretching the bow tie’s weight handle changes the degree of its influence, and pulling on the curvature handle changes the radius of the spline’s curve on either side of the bow-tied point.  These adjustments can be slippery, which is why it’s best to use dimensions to tweak bow-tie handles.


Lofted surfaces can act somewhat like 3-D splines.  Thus the same rule applies:  the fewer sketches in a lofted feature, the better.  Rails are a useful tool for smoothing out lofts, but I avoid them if I can because they make edits cumbersome.  Projected work points coincident to both sketch and intersecting rail make good hook-up points. 


The “conditions” tab on the loft feature is vaguely similar to a 3-D bow tie.  Setting the weight and tangency angle of this option changes the influence of a loft’s first and last sketches, making two-sketch curved lofted features possible in the same way bow ties make possible two-point curved lines.  The conditions tab is especially useful for floating a lofted surface smoothly off another surface, again similar to the action of a tangent bow tie.  By the way, said mating surface can serve as the first or last loft cross section; a loft doesn’t have to terminate with a sketch.  3-D sketches and undulating surfaces are also suitable. 


This example contains no sketches whatsoever.  Two cylinders’ opposing faces served as terminators and conditions determined the curvature.  If the conditions are given more weight, this shape results from the same sketches.


One can’t loft to a point, and lines can terminate surface lofts only (not solid lofts).  But there’s an easy cheat.  For a point, make a circle sketch and dimension its diameter to .005.  Loft to that sketch and then fillet the resulting feature’s microscopic blunt end to R .0025.  Same trick for a line.  Using conditions, this mirrored example lofts from previously extruded features to a microscopic circle sketch.  Shawn Clemment used the same trick to put a point on a model of a sword.


Attempts to loft complex sketches will often result in error messages.  When that happens lofting as a surface rather than a solid will often solve the problem.  Use the lofted surface to slice the desired solid shape from an extruded block.  Another trick is to break the sketch down into segments which can then be lofted one piece at a time using “share sketch”.    By the way, most consumed sketches can be re-used even if no share-sketch option is available.  Simply turn on the sketch’s visibility and re-select it for the next operation.


I generally use two basic frames when setting up a loft.  I either line the sketches up along the feature’s length, or I distribute them about a center line, like the sections of an orange.  The race cars shown on this site were lofted using the first method, the sports cars using the second.


Splines are easy to adjust on the fly.  When designing a complex lofted form like an auto body, one wishes there were a way to adjust 3-D features by similarly pulling on points.  It turns out you can:


Turn on the visibility of all the feature’s sketches.  Pull on any sketch point not constrained by dimensions or connected to a rail.  The sketch will stretch free from the feature’s surface.  Now hit “update” in the top toolbar.  The feature surface will jump to match the new sketch contour.  For a symmetric shape like a car body, it’s useful to use mirrored, undimensioned sketches so that changes will be automatically reflected on the side opposite the edit.  Bow ties horizontally constrained to each sketch’s mirror points will prevent peaks or clefts from occurring at the feature’s spine—which by the way should be a common work plane projected into each sketch to keep the feature in line.


This stretch-and-update method is more difficult where sketch points are constrained to rail points.  So far I’ve found only a partial solution: 


In the upper tool bar, highlight “select sketch elements.”  Left click on a sketch.  Release the mouse button and move off the sketch, then return to the sketch and grab its end point.  The point can now be pulled away from the surface, leaving the previously connected rail endpoint behind.  Hit “update” and both the feature surface and the rail should jump to the new sketch position.  The rail endpoint can be adjusted the same way.  I have yet to discover a means to stretch coincident sketch/rail points other than endpoints except by disconnecting the points in “edit sketch” mode, a tedious process.  Swirly.ipt shows a railed loft before and after stretching.  When using this technique it’s fun to use the forward and back arrows to compare your part’s contours before and after the stretch:  Do I like this better, or this?


Pillows” illustrates a hazard to loft and similar surface-modeling features that can sometimes be used to advantage.  The “variable fillet” feature was used to round one edge of one of a pillow’s elliptical ends.  Since fillets don’t mirror, the filleted quarter of the pillow was then imported into a derived component file and mirrored as a part.  The resulting half-pillow was again derived and mirrored to make a whole pillow with all edges filleted.  The aforementioned hazard—in this case a benefit--shows up as rumpling.  Rumples are caused by excessive pressure where feature bends are over-crowded.  A close look at the race cars’ front fenders will reveal tiny wrinkles due to my use of an angled sketch at that point (a decided mistake) but when modeling fabric, skin, fluids, etc., rumpling can be useful.  The variable-fillet/mirrored derived-part technique was suggested by Eric Hall.


The “wing” files illustrate the massive economizing capabilities of surface modeling features used in sequence.  Here a hollow airplane-wing blank, including brace and notch cross-sections, was lofted piecemeal and then sliced into ribs via intersecting patterned extrusions.  Spar sketches were then projected into an assembly and lofted in place.  The corresponding spar notches were then projected and cut-extruded.  Finally, the wing’s skin was lofted over the frame using an offset sketch.  Some of the sketches used here were streamlined by converting projected curves to normal style to allow trimming.  Simply select the projection and click normal in the “styles” drop-down in the toolbar.


Both parts of “Wiggly split” were similarly carved from a single solid loft.


In future I plan to update this site with notes on other features including “sweep”, “thicken,” “delete face,” “remove lump,” “shell,” etc.  In the end it appears the conclusion of my investigation will be that Inventor’s limitations as a surface modeling tool are surprisingly few.















Blank Page 5

Blank Page 6

Blank Page 7

Blank Page 8