Wheel and Axle "Kit" second attempt

     After revision of dimensions of the axle diameter and axle hole in the driver wheels, a 2nd run was instituted.  Because the axle was so difficult to remove from the raft and supports, it was elevated above the raft in an attempt to obtain an easier release.  The view at left shows the kit immediately after printing while still on the printer Z platform.

     After removal from the printer the raft under the wheels and the thick support structure under the axle are clearly visible.  The raft and support are clear PLA while the wheels and axle are black PLA.  It took 6 hr and 32 minutes to build this set of parts.  At first impression the top surface was about equivalent in finish as the first run, but somewhat more strings of black material were evident.  The strings are not a problem as they can easily be removed.

    Raft and support removal was attempted as before using finger pressure or a needle nose pliers.  The thick support under the axle readily came off with finger pressure leaving the axle nearly free of clear PLA material.  The wheels were another story, however.  The clear PLA maintained a tenacious hold to the wheel and rod crank areas.  Removal resisted extreme measures using both a needle nose and side cutter pliers.  A clear layer of PLA is clearly shown in the view at left.  The axle however does seem to fit much better than previously.  Measurements will still need to be made to verify that both parts approach their design accuracy.  Another run will be attempted using a much thicker raft and support under the wheels as well as the axle in the belief that it may be easier to remove all parts from the raft and support base.

Axle post build processing results

    Both the end up and diameter down axles were post processed using a hand file and Dremel tool sanding wheel.  Both were improved in diameter smoothness and consistency.  The end up version raft end bulge was reduced from an oversize condition of 0.045" to 0.021".  It was hard to judge when the axle was uniform, however, it should be possible to improve it more by making measurements during the effort.  The top end was not processed and remained oversize by 0.010".
    The diameter down axle version initially was oversize by 0.0265" and 0.0207".  The raft attachment area was very rough due to removal using a side cutter pliers.  The file was used to make the attach point smoother and more uniform.  This was followed by use of the Dremel tool sanding wheel to improve roundness along the attachment area.  The final result was 0.0131" and 0.0142", considerably a better match than was obtained with the end down axle.
    In recap, the axle holes in the two driver wheels were 0.0173" and 0.0187" undersized.  The combination results in an interference of approximately 0.04" using the end down axle and 0.033" using the diameter down method.  Since the diameter down axle results in better uniformity between the ends, it appears that the Wheel-Axle Kit would benefit by reducing the axle diameter about 0.0176" and increasing the wheel axle holes by 0.022".  This should result in an axle diameter closer to target of 0.0412" with a clearance between the wheel hole and axle diameter of a few thousandths of an inch.  Careful post processing should result in a reasonable slide fit suited for bonding using glue.
    The driver bearing parts will also need to have their axle hole adjusted to clear the axle sufficiently to permit easy rotation without binding.   However, due to the extensive amount of support required, it would be better to build the bearings using the ABS raft and support which is more easily removed using needle nose pliers or fingers.

Issues in building locomotive axles

    The axle built during fabrication of the wheel axle kit run exhibits a noticeable bulge and roughness at the bottom end where the axle came into contact with the raft and support.  The support was necessary for  a small overhang due to presence of the keyway indentation at the bottom.  Another keyway is at the top however it had a floor instead of an overhang and almost no distortion was present.  The majority of the axle diameter is around 0.422" while the bulge was 0.457".   The design target is 0.412",

    Another axle was built laying down along the diameter using PLA raft and support with PLA build material.  A narrow layer of clear PLA raft material proved much more difficult to remove requiring side cutter and other tools.  After removal a narrow flat region exists on the portion of the diameter where the axle came into contact with the raft.  The diameter along the axle was more consistent and did not exhibit the bulge seen during the previous build.  The diameter was approximately 0.43" over the entire length.  The diameter is still oversize from the design diameter of 0.412".  It is probably worth building another axle with ABS raft material since it seems to separate more readily from PLA than does the clear PLA raft material from the black PLA build material.
    These two parts will also provide test cases for post build processing to improve the uniformity of the axle diameter.  This will probably be accomplished using a Dremel tool.  No attempt to reduce the overall diameter to the design diameter, only to make the part more uniform in diameter.  Since both the axle and wheel axle hole are off in opposite directions, the axles will not fit into the wheel at all.  Consequently, the design will need to be revised to enlarge the wheel hole target dimension and the axle target diameter.  By taking into account the relative errors reported previously, it should be possible to adjust the part build dimensions so that the axle will fit into the wheel axle holes.  

Building a PLA model on PLA raft and supports

     A PLA model was built using PLA raft and supports.  The wheel axle kit was used which includes a right side and left side driver and the interconnecting axle.  At left the picture shows the model upon completion in the BFB-3000 3D printer. Natural (translucent clear) PLA was used for the raft and supports while black PLA was used for the model proper.  The build cycle was 6 hr 33 minutes.  The axle is standing on one end between the two wheels.

    After removal from the printer the supports appear nearly transparent as shown at left.  Initial inspection indicated good resolution of both wheel flanges and good finish and surface continuity, much better than was obtained previously using a PLA raft with ABS model material.

     The model parts were readily removed from the raft and supports, generally more readily done using a needle nose pliers.  Wheels and axle at left are shown alongside the removed raft and support materials.  Some black PLA thin flash materials were also found and easily removed.

    At left is one wheel photo made using a flash.  Certain edges strongly reflected light.  The overal wheel finish is quite smooth and generally semi-gloss black.

    The axle was quite uniform in diameter except near the bottom where the keying feature required support material.  The axle in that region appears bulged.  The axle will require post processing to obtain a uniform diameter throughout it's length.  

    The other wheel photo was also taken with flash lighting and again has reflections due to the glossy surfaces.
    The next step will be to measure the parts and determine how accurately the parts conform to the design dimensions.  A preliminary fit check revealed that the axle will not fit into the wheels unlike the ABS parts.  The ABS wheel axle kit run resulted in an undersized axle enough to permit a fair fit on one wheel and a very loose fit on the other.  The axle will ultimately be bonded to the wheels, however, the fit should be close to preclude misalignment as much as possible.

Accuracy Analysis of PLA Bearing Part

    Each driver wheel has a bearing that rides up and down in a slot of the chassis frame.  The bearing surrounds the axle near each wheel.  On top of the bearing is mounted a spring connecting plate that couples the load from the driver to the chassis through a set of equalization levers.
    Four copies where made during the Small Parts Kit build.  The picture at left shows the parts which have been numbered for the accuracy analysis.

    The picture at left shows the bearing parts next to a ruler for size reference.  The bearings are 0.81" X 0.95" width and length.  Each has an axle hole of 0.42" target dimension.

    The table above provides a comparison of design dimensions for key features vs. measured values averaged over the four parts.  Statistics were also computed that measure the dispersion (sigma) and likely range of values (Error +/- 3.3*sigma) that may occur for other copies of the bearing that may be built in future.  As can be seen the greatest error arose building the axle holes.  Those are noticeably undersized.  The next largest error arose with the slot that interfaces with the chassis.  The slot span side to side is oversize by a bit over 5%.  Other key dimensions are also in error as shown.  In order to compensate for the errors either post processing (drilling, sanding, etc.) would need be performed, or alternatively, modify the design to adjust those dimensions in the opposite direction in an attempt to obtain better accuracy.  Perhaps both will be needed for some parts that are expected to move in respect to one another as will be the case for the bearing in which  the driver axle rotates and the bearing itself moves up and down in the chassis as the equalization scheme adjusts for small irregularities in the track, as is done on the prototype.

The chart above summarizes the nominal and predicted range of possible errors based on the statistics of measured values for key features of the four bearing assemblies.  The green squares are the average errors while the red bars depict the range of possible errors based on the application of sigma variance to the averages.  Corrections to the design dimensions should reduce and center the error bars around zero.  However, several of the dimensions have rather large error bars which might prove difficult to use without considering post machining of the parts.  On the other hand, it is not always necessary to obtain tight fit tolerances.  The primary objective will be to ensure parts stay together and those that move do so without slipping off or out of position.  In the end it will be necessary to build and fit check actual parts, not just designs.

    The preceding chart shows the trend of bearing model errors vs. size of feature.  It appears that feature sizes of approximately 1/2" are most accurate, those having lower dimensions tend to be undersized and those greater tend to be oversize.   Charts such as these will aid in developing methods to obtain suitable accuracy for critical features that must interface with appropriate precision, such as moving parts.

Small Parts Test Kit

    A test assembly "kit" was made up of various part models and printed in one run.  Due to the overall size and complexity the run was a bit over 8 hrs.  Starting at the top in the left view the two "C" shaped parts are trailing truck spring saddles, between them at the top are four small front saddle mounts.  Below those are four rod shaped front trailing truck spring pull rods.  The two rectangles at the left with an "X" in the middle are frame spacers used on the front and middle frame assemblies.  At the right of those are a pair of spring "A" and "B" parts, "A" at left.  Below the frame spacers are two brake hangers that go on the front and middle frame assemblies.  Below those are four bearing assemblies for the front and back driver axles that interface with the frame and driver springs.  At the right of those are another set of trailing truck spring "A" and "B", "A" at left.  Surrounding and below the parts in translucent clear PLA is the green ABS raft and support structure.  Much of the time was spent building the ABS portion.  That portion is sacrificial and facilitates build and support during build of the PLA parts.

    At left is a close up of the four bearing parts that clearly shows the surrounding green ABS supports and the green ABS raft below.

    The two frame spacers did not require supports for overhangs and rest directly on the green ABS raft.

    The raft and supports did not always build properly.  At left is shown one of the two trailing truck spring saddles and above left one of the four front mount for the saddle is lifted.  The saddle is separated and distorted.  Above the right side of the saddle are three of the trailing truck spring front pull rods.

    The four pull rods are shown in position on the raft at left.  Some ABS supports were used to make the build.

    Two of the four front saddle mounts are distorted. This is due to the delamination of the raft structure.  The information provided by the manufacturer indicate that ABS will distort when structures over 100mm are built.  That is the case here, so the raft delamination is somewhat expected.

    The picture at left shows the other "C" shaped trailing  truck spring saddle.  This one came out good.  The green ABS material nearly surrounds portions of the model, however, it was readily removed by pulling the material away using fingers.

    Spring parts "A" and "B" came out fairly well, however, insufficient separation between the leaves resulted to permit the two structures to be interleaved.  Both portions operated as springs quite well and it appears that a design change using a modified "A" version will work quite well.  The "A" spring is at top with the round pin mount at left.

    The other spring "A" section did not come out well.  The PLA material was spread around near the top and leaves were distorted.

    The view at left shows the good spring saddle and saddle front mounts and pull rods above it.

    At left is another view of the four front-back driver bearing assemblies illustrating the extensive amount of support material used to deal with the overhangs present on the model.

    The driver brake hangers that attach to the front and middle frames are shown at left middle.  They also have extensive amounts of ABS support structure surrounding the parts.

    The view at left shows the parts at right after separation and at left the residue of the support and raft structure.  The green material is ABS while the white translucent parts are PLA.

    Overall the results were fairly promising indicating that small parts can be built successfully, particularly in PLA.  The small front saddle mounts are about 1/4" X 1/2" and have six bolt details, which are present.  The detail fidelity of the bolts is not very high, but they are recognizable as round objects and in the minds eye they are either bolts or rivets.
    The dimensional tolerance is similar to other models previously made, IDs tend to be undersized as do some outer dimensions.  Some are over sized, particularly outer dimensions of small objects.  These parts will be measured and compared to design values as part of an accuracy study.  The results should provide rules of thumb for size adjustment of design dimensions to result in somewhat better fit between mating parts.

Cleaning and Assembling the Driver "Kit"

    The Axon conversion software adds support structure to prevent overhangs from sagging or collapsing.  A small overhang occurs on the axle and the counterweight portion of each driver.  The first step is to peel away the primary raft of PLA material and the supports.

    The separated parts are shown sitting on the raft at left.

    A close up view shows the remains of the white support PLA material.  It is easily removed using tweezers or needle nose pliers and may be trimmed with a model knife if need be.  The bottom of the drivers shows remnants of the raft.  This is a thin layer and can be removed with sandpaper which also smooths the raft interface surface.

    The parts fit together readily as shown at left.  They will later be bonded together with some form of epoxy or super glue during the assembly process.  Two additional parts, the bearings, go on the axle to interface with the chassis and springs.  They are much smaller and have yet to be printed.
    The wheels and axle will need to be painted to the final color before assembly.  This group of parts forms an assembly with 2.675" diameter wheels on 2.875" centers approximately for size reference.  The rail heads will be 2.354" across the inside edges.

Wheel and Axle "Kit"

    A test of part group printing was made using a set of driver wheels and the corresponding axle.  Green ABS was used (did not have the correct black on hand).  The parts are shown at completion on the printer bed prior to removal.
    The white grid at the bottom is the PLA raft that provides a flat surface for printing the ABS parts.  The axle between the wheels has some PLA support material near the bottom due to the presence of a keying indentation that required support material.  The wheels also have some support material particularly under the counterweight portion between the spokes and under the spokes as those portions require support to prevent the ABS from sagging while hot.

    The photo at left shows the ensemble of parts located on the printer bed of the BFB-3000 3D printer.  The bed is in the lower position to provide access for model removal.

Impact of hollow axles

    In an earlier entry it was postulated that hollow design for the axles and perhaps other parts would result in faster printing.  The left hand axle shown is solid while the right hand one is hollow per the design.  The results of hollowing for the rather thin wall of 0.05" is splitting of the wall at the point where the Axon conversion software reverses from clockwise to counterclockwise following of the wall.  Both axles were printed upright as shown.  The end portions of the right hand model have a tapered fill inside that gradually closes the hollow so that the end caps are solid.  At some thickness of the taper the split discontinues, consequently the wall thickness is likely the real issue.  Evidently 0.05" is too little wall thickness and needs to be greater.  The solid axle at the left has a 50% fill factor selected during the Axon software conversion of the design and is actually not fully solid either.  However, the walls are much stronger due to the presence of fill material (also ABS) that provide reinforcement for the walls.

More initial results with BFB-3000 printer

    After printing the factory test pattern (a small duck model) a locomotive driver wheel was printed using the same set-up, ie, white ABS for the raft and fill and black PLA for the model.  The 3D CAD model is fairly complicated with 421 faces, 1025 edges and 603 vertices.   The model is 2.66 inches maximum diameter and 0.52 inches thick.

    The 3D model printed on the BFB-3000 was good with the exception of the flange.  The flange is quite thin and located on the lowest layers.  Apparently subsequent printing of the tire portion of the wheel injected sufficient heat to distort the flange.

    The wheel and factory duck model are shown with a ruler for size.  The duck model is quite smooth as is the wheel.  The duck had a couple of very small imperfections under the chin of the duck.  As was mentioned the wheel flange is badly distorted.  The white layer under the wheel is the raft of white ABS plastic.  It is composed of alternating X and Y grids to provide a level surface and good attachment interface between the build table of the machine and the model.

    Another Axon file was created defining the same wheel model be built in reversed materials.  The raft will be black PLA while the model will be white ABS.  both models are shown at left.  This time the flange came out good.  The density of internal fill was reduced for the white version which resulted in a few internal voids, one can be seen on the right side.  Future ABS runs will call for a higher fill density or alternate fill pattern to reduce or eliminate the voids.

    Measurements made of the wheel indicated some error of dimensions.  To study the error possibilities a line and space test pattern was prepared using the Alibre 3D CAD software to evaluate lines and spaces from 0.100 inches down to 0.01 inches.  The design is shown at left.  Each line and space combination was printed twice in both the X and Y axis forming two "L"s of lines with the target dimension width and space.  The next lower size is spaced the same amount giving two spaces for evaluation.  The dimensions are 0.1", 0.08", 0.06", 0.05", 0.04", 0.03", 0.025", 0.020", 0.015", 0.01" and 0.008".  The printer information indicates that vertical walls such as used in the model must be at least 0.025" thick, so the lower dimensions are expected to have problems.

    The resulting print with an ABS raft and using PLA material for the model is shown at left.  The print appears quite good for the three larger lines of 0.1, 0.08, and 0.06 inches.  The next lines at 0.05", 0.04", 0.03", 0.025" and finally 0.02".  The Axon software eliminated the smaller dimensional lines of the 3D model.  All lines are the same 0.25" height.  Overall the model looks very good, but did exhibit various dimensional errors.

    The blue line on the graph at left shows the ideal dimensions for the lines measured while the red dots show the actual value averages for four measurements at various positions on the lines.  The dotted line is a curve fit showing the trend of the red dots.  It is evident that as the line width increases the dimension develops an error of up to 0.012" or so while the smaller lines approach correct dimensions.

    The graph at left depicts the space measurements vs the correct values.  Once again there are errors which are most noticeable for middle size spaces around 0.05".  Spaces at smaller dimensions are somewhat better while at the largest spaces the size is actually less that the design dimension.  These errors are all for relatively small dimensions of 0.1" and below.

    A large test grid design was prepared to evaluate the errors for very large dimensions up to 9".  The pattern consists of squares of 1" X 1" with internal squares of 0.5" X 0.5" with some of those having internal squares of 0.25" X 0.25".    The Axon software was again used with ABS raft and PLA model material.

    The large grid pattern is shown at left with a ruler used to check for dimension accuracy.  The spacing between inch marks on the ruler exactly matched the position of the 1" X 1" grid bars in both dimensions.  Also, the 0.5" X 0.5" bars were similarly accurate in spacing.  Once again the bar widths exhibited errors similar to those observed with the line and space test pattern.  This pattern is near the largest X and Y dimensions for the printer and provides a good check of large part fidelity.

First Print Results With BFB-3000 3D Printer

     The first models made on the BFB-3000 are shown at left.  The small duck model is a test file provided by Bits From Bytes to verify the correct set-up of the printer.  The locomotive wheel on the right is one of the drivers built to 1/2" per foot scale.  The prototype driver is 63" in diameter, the model is 2.625" not including the wheel flange.






    The flange on the locomotive wheel is badly distorted.  The reason is not clear but could be the heat of subsequent layers of molten plastic being applied along the tire rim near the flange.  The white plastic used for the built support is ABS while the wheel is PLA.
    In order to determine the better ways of building parts, some test parts will be designed and built to determine the best arrangement of layout and the appropriate sequence of plastics to use.
    Each sequence consists of a support "raft" and overhang support plastic and the primary build plastic.  Four combinations are possible.  All PLA, All ABS, ABS raft and PLA build (as was done for these first two models per vendors set-up procedures), and PLA raft and supports with ABS build.  Some designs having little or no overhangs can be built without supports.

3D Printer Arrives

    The BFB-3000 3D printer arrived this week.  It weighs 70.5 lbs and will require help to remove from the carton.  It is about 20" on a side with quite a bit of metal internal structure and smoked plastic outer covering.  It is shown in the shipping box at right.  It was packed with custom foam inserts to support and separate the parts.  Not shown are the two top foam inserts and the plastic top for the machine.  Much of the internal workings are secured in place with plastic tie-wraps and foam wrappings that will need to be removed.
    The unpacking instructions point to a web site where the set-up and operating procedures are found.  They can be downloaded to a local pdf file on my computer to provide a ready reference in future.  A preview of the instructions revealed that it is not recommended to build ABS parts over 100mm as they tend to warp.  This will likely change the selection of primary materials to PLA where long parts are needed.

    A small storage rack was found at Lowes and the lower half of it provides a strong little stand for the printer.  The printer comes with a selection of PLA and ABS material on spools each in a separate cardboard box.  The shelves of the rack provide a convenient location for spool storage until needed.  The rack is being located next to the computer desk.
Once the printer is in place and set up, various objects will be printed from the design to ascertain the true capabilities of the printer and materials.
Alibre (http://www.alibre.com/3dprinters/default.asp) now sells several lower cost 3D printers, the least costly was just introduced, the BotMill which is listed at $1395 less the Alibre design software or $1494 with.  Alibre also offers the RapMan 3D printer kit (very similar to the BFB-3000), the BFB-3000 and the V-Flash (listed at $14,999!).  The BFB-3000 was selected for our projects seemingly to provide an affordable higher quality printing capability.   The BFB-3000 was ordered directly from the manufacturer, Bits From Bytes in the UK, alternatively it can be ordered through Alibre Design.