Lathe Design
Project Background
The most in-depth modeling and design project that I have worked on thus far has definitely been my lathe, into which I have invested over 1000 hours of my time modeling, searching for and acquiring materials, and solving the relevant math problems involved in producing a working machine. Originally, I wanted to design and build a simple lathe just to turn wood with, since I’d seen videos online and it seemed like a satisfying hobby. I don’t like purchasing things when I can build them myself, though, and at the time that I came up with the idea, I had about a year and a half of 3D modeling experience under my belt, so I got to work designing my own.
I wasn’t intending to turn anything large on my lathe, and I wasn’t concerned much with precision, so I settled on building a wooden frame with imbedded ball-bearings. To act as a part-driving mechanism, I paired a ¾-inch rotor that I had pulled out of a washing machine together with a two-horsepower motor that I found used online. I never did much research into how wood lathes were normally structured, so my first design mixed some concepts drawn from both wood and metal lathes that I had seen as well as things that seemed intuitive to include.
As I searched the internet for answers to the design questions that I had while preparing my lathe model, I found many examples of metal lathes and their applications which I was far more impressed with than the imprecise and primarily manual capabilities of wood lathes. I found a few lathe-reconstruction and home-built lathe project videos online that utilized aluminum-casting and steel-welding to form the main structural base of their lathes. Since I had just finished my foundry and had collected a large amount of aluminum, this seemed to me as though it would be a feasible and far more structurally sound option for my lathe.
I became more interested in making a solid lathe capable of machining a large range of materials rather than quickly throwing together something that would barely get the job done, and I was willing to actually invest real money into it as a new in-depth project. My respect for precision instruments and machinery grew rapidly in the following months as I continued to research the materials, tools, and processes that I would need to use in the process of building my lathe.
I spent nearly a thousand hours of my spare time over the course of around a year modifying and adding components to my lathe assembly and designing fastening methods and easily manufacturable webbing to make the lathe as rigid a structure as possible. I began the project expecting that wood would be a sufficient base material, then moved to aluminum casting, then eventually moved on even from that to welding a structural steel frame, deciding that the strength and manufacturability benefits outweighed the literal weight that the frame added to the system, from a practical standpoint.
As it stands today, I have modeled 102 separate parts, all used in the final lathe assembly, and many used multiple times. The entire assembly is made up of 591 total parts, mostly fasteners, all in a mechanism built on the motion constraints that a real-world version of the lathe would be bound by. I have purchased many of the materials from which I will construct my lathe leaving out only the drive-train and quick-change gearbox components which I will work on after the rest of the lathe is assembled. The next step in my building process is to use my modified band saw to cut the scrap steel that I intend to form the lathe’s structure from into the necessary constituent pieces before welding them together and grinding them to precise final dimensions.
Despite the majority of this project being CAD modeling and far less of it being real-world machining and construction thus far, I very much enjoyed all of the work that went into it. Through this project I have discovered that I eventually hope to design and prototype precise mechanisms through use of CAD in my career. This lathe turned into such a complex machine, as it should be, and I am very proud of the work that brought it to its current state. I look forward to the machining and assembly that will be necessary to move my lathe from 3D project files into the real world, and I hope to learn more along the way about the processes best suited for that kind of work.
Detailed Description
When I began modeling the Treasure Cube, I wished to improve as many of the official Treasure Cube’s shortcomings as possible to decrease the chance that my cube could be beaten by any method of opening it, through brute force or otherwise, besides solving it. Normal cubes are weak enough, with high stress concentrations occurring at internal corners that appear everywhere within the structure, but the design for the Treasure Cube does have a bit more structure to it than the conventional cube model. It supports each piece with a track rather than through the center faces being attached to a six-axis core, which leaves the entire center cavity completely open and available as a container in which to store items.
The modeling process began with the dimensioning of a base structure on which the lathe components were to be mounted, which I call the lathe bed. In my first model, I intended to use wood to form that bed, so I used standard wood-stock dimensions to drive the dimensions of the parts in my lathe bed model. I planned to bolt multiple pieces of wood to each other to form a complex 3D structure which, I expected, would be able to handle the loads that the lathe would be subjected to. I included a location for the motor to be mounted after modeling the motor itself with dimensions that I pulled from the real motor once I had it in hand, two vertically protruding flanges on which to mount the bearings, and a single flange at the end which I intended to use as a fixed tailstock before I considered that they are intended to move.
Wood-lathes don’t have saddles and therefore don’t need leadscrews, but I added a leadscrew and linear rails for something of a saddle to ride on, mostly because I had not yet begun any real lathe research. The second mistake that I made before doing my research was becoming emotionally attached to and purchasing a cheap self-centering three-jaw chuck that I found online. Using schematics for the chuck given on its manufacturer’s website and dimensions that I pulled from the chuck itself once it was delivered, I modeled it in Creo and attached it to a backplate mounted on the spindle that I had added.
Another major difference between my design and professional lathe designs was that I directly drove the leadscrew in my original model with a wheel which required that the wheel be located at the end of the lathe, making it difficult to access during usage. I have mentally marked this basic design as my first attempt at modeling the lathe, and I count every subsequent model adjustment as an improvement. I had to start somewhere, but I had no idea just how much I would end up modifying before I was satisfied.
My first adjustment was one to improve the rigidity of the system, in which I replaced the end-supported linear rails that I had modeled with fully supported, bottom-mounted linear rails. This way, the flexing in the beams caused by the extreme cutting forces and the weight of the carriage would be minimized as the forces would be transmitted instead directly into the surface that the lathe sits on.
Rather than mounting the bearings supporting the rotor directly in the wood, essentially forming a pillow block bearing, I opted to purchase cast iron pillow block bearings to mount atop the wood flanges, adding strength to the system and a better way to handle any axial forces that the workpiece may experience. I then modeled the pulleys that would transmit motion between my motor and rotor axis, not realizing that the cutting speeds that I would get from this setup were far higher than those desirable in driving a lathe, which I eventually fixed later on with the inclusion of a gearbox. The pulleys have five available settings giving me a 3:1 ratio, 5:3 ratio, 1:1 ratio, 3:5 ratio and 1:3 ratio between the motor and lathe spindle speeds at those settings.
My newfound interest in lathes led me to search online for others working on home-built lathes to give me ideas and allow me to consider their mistakes before making them myself. One major improvement that I realized that I should make was in the bed, as I had drastically overestimated the rigidity that I would achieve through one made of wood, and with such a simple structure based only on planes of wood and few fasteners with no real bracing. I also wanted to prevent the frame from warping in any way over time which could only be done by swapping out the wood with metal.
I entirely scrapped my wood frame idea and to replace it I designed a new lathe bed and motor housing, both constructed as one section to be cast as a single object. This, again, was a starting point for many improvements to come. I expected that I would make major adjustments down the line in terms of the manufacturability of the bed, material usage, and the structural distribution of that material.
Having essentially started from scratch, I reassembled my components from my old lathe model to my new bed structure, this time using four separate bearings to support the spindle shaft. I decided that this was necessary because, at only a 0.75-inch diameter, the shaft was prone to flexing between two points which would throw off the axial aim of the chuck. Therefore, since I knew that constraining two more points would largely resist that bending because of the added nodes, and because they were cheap, I implemented four equally spaced pillow block bearings atop the motor housing.
I had realized the error in leaving the saddle-drive wheel at the end of the lathe, so I instead moved it to the side with the intention of gearing it to a lead screw. Most lathes that I have seen use a lead screw to drive the cross slide but use a rack and pinion to drive the saddle. From the start, I did not want to do this with my saddle, and I went with a ball screw because I thought that I could perform far-more-precise saddle movements with a ball screw than a rack and could eliminate the backlash problem inherent in acme screws and rack-and-pinion systems.
Realizing that a single cast aluminum frame would be very difficult to manufacture, especially one with the aesthetic contours that I added in my model, I again designed a new frame and tried to obtain a similar structure using far less material. In this new version, I separated the motor housing and the ways section, or the part that carries the saddle consisting in my lathe of the linear rails, to be cast as separate parts and mounted to one another via fasteners. This design ended up more rigid than the previous despite the great difference in material usage due mainly to my inclusion of easy-to-cast hollow sections beneath the bed.
Using a combination of fairly simple geometries, I would be able to cast this new base after preforming those sections from wood and using greensand to transfer their shapes to a negative cavity. I had plenty of aluminum scrap available, mostly in aluminum cans and some in flat, machined aluminum sections taken from a retired spectroscopy optics set.
As I continued to model new components resembling more and more those found in real lathes, I realized that I needed a proper saddle and cross-axis slide. My earlier models utilized my scrap aluminum flats as the structural sections of the saddle and cross-slide to which linear bearings and rails were mounted. I had purchased a mill vise, so I modeled it as well and placed it on top of my cross-slide, hoping to use it essentially as a tool-post, which I’m glad to say I eventually replaced with a real tool post.
Beneath the aluminum saddle sheet, I included four linear bearings to ride on the ways linear rails prepared in a square pattern. On top of the saddle, I installed two more linear rails perpendicular to those below whose travel would be parallel to the lathe bed. I mounted the vise atop this second stage of motion to give the lathe the ability to change the diameter of the cut in addition to the location of the cut along the workpiece.
As a solution to the error in alignment that I knew would exist at first between the spindle axis and the direction of saddle motion, I mounted the four bearings supporting the spindle to a separate steel plate instead of directly mounting them to the top of the motor housing. This plate would keep them rigidly aligned with respect to each other and was mounted to the top of the motor housing using eight evenly-spaced bolts. The bolts in the corners would allow fine control of the spindle axis direction and when all eight bolts were tightened, they would provide the necessary rigidity for retaining the spindle in that position despite the loads that it would be subjected to.
The width dimension that I had originally selected for the lathe bed didn’t sit right with me, especially since there was so much overhang in the saddle, which I knew would translate to large moments inflicted on my saddle’s linear bearings under large loading conditions when applied to the extremes of the saddle. To remedy this, I widened the base by a factor of 1.5 from six to nearly ten inches across. At this point I decided to re-introduce my ball screw which I hadn’t yet tried implementing into the new aluminum base, then modeled and installed the pillow block bearings sold with it to constrain the ballscrew at its ends. With the new large increase in the width of the lathe, I felt that it would be okay to increase the length of the rails on the cross-slide by a total of 10 centimeters, (since they are sold in intervals of 10 centimeters), giving me more usable cross-slide travel.
At around this point I began modeling my tail stock, this time mounting it on linear bearings rather than fixing it at the end of the lathe. My first model of the tail stock was made after I made my foolish design choice to use two long-type linear bearings instead of four distributed linear bearings for it to ride on. I thought that this would provide enough strength to support the tail stock and its expected loads, but as I investigated, I saw that the moments that it would be subjected to were to be far too large for the bearings to rigidly support the tail stock without risking permanent damage to the balls because of the small length of ball distribution within the bearing. I originally wanted to use these bearings to save a bit of money, but the massive loss in structural soundness that they would cause made me reconsider and later switch to four, small, more evenly spread bearings.
While investigating the limitations of the linear rails that I had selected, I discovered that profile rails are superior to cylindrical rails in strength and bending resistance and are constrained at far more spline points than cheap cylindrical rails, making their connection to the base more rigid. I swapped out the cylindrical rails in my model for profile linear rail guides both to reduce the total height between the bed and spindle and because of the many benefits of using profile rails, especially their improved load ratings and resistance to torsion and forces applied upwards.
The decision to keep the lathe this way, using more expensive but superior profile rails, or to switch back to cylindrical rails was a difficult one because the upside of the new rails was their improved mechanics, but their real downside was that they were nearly twice the price of similarly sized cylindrical rails. I did end up switching back to cylindrical rails as I believed that I could overcome their weaknesses by bolstering other components and modifying the rails slightly.
I imported sections of the cylindrical rail into a finite element analysis software called Ansys, which I had been using in school, and ran analyses on them with the loads and boundary conditions that I expected them to experience within the lathe. The displacements reported by my analysis had me convinced that the cylindrical rails would be sufficient for my needs in the lathe bed and cross-slide and that the only thing that I needed to worry about was reaching the load rating of the bearings themselves, rather than the deflection of the rail sections caused by cutting forces.
I didn’t much like my original idea of mounting the spindle bearings on a separate plate, so I removed the plate altogether and directly mounted the bearings to the motor housing which both lowered the spindle and reduced the number of fasteners. To solve the problem of spindle-axis alignment, I allowed for some play in the mounting holes and installed 5/8-inch locknuts on either side of the pillow block flanges to finely control their heights without the unnecessary intermediate steel plate.
In my first attempt to model a functional tail stock, I made its whole frame as one part which, when translated into the real world, would mean that it would need to be cast as one section. I didn’t necessarily intend for it to remain this way and planned to later change its design so that the tail stock structure alone was cast from aluminum. That structure would then be attached to an intermediate plate of aluminum or steel mounted atop the linear bearings, so I wouldn’t need to essentially cast and machine my own baseplate.
Up until this point I had used a model of the motor housing entirely constructed from and to be cast as one part. The difficulty with which my motor and bearing mounting holes would need to be drilled into that housing made it a very impractical design, so once again, I remodeled it from scratch. My new version had essentially the same dimensions, but it was designed to be more rigid, more easily manufacturable both in the drilling and casting sense because of its simpler geometries, and to be made from three separately cast sections to be connected to each other later on. It now consisted of a front panel, back panel and mid-section.
Evenly-spaced large-diameter machine-screw holes were added with tapped holes only in the mid-section. I planned to mill the connection planes between the three sections to be flat and then drill, ream and pin some locating holes once the three components were aligned in their final assembled state so that I could reassemble it exactly the same way if I needed to perform maintenance on any component.
I then made my first adjustments to my tail stock model, bringing down the diameter of the outer encasement to cut down on wasted material and eliminating the aluminum baseplate that I had included which I replaced with a one-inch-thick steel plate. The two legs of the tail stock structure were modified to have a flange between them at their base through which bolts would fasten the flange to the steel plate. I also replaced the two large tail stock linear bearings at this point with four smaller ones, as I should have done from the start, which increased its footprint and its capacity to resist bending moments.
As the basic structural components had been modeled, I moved on to modeling the supporting and driving mechanical components like machine screws, nuts and bolts, wheels, timing belts and pulleys, axial bearings, ball screws and their respective nuts. I created a new wheel to be used on the tail stock and cross-axis lead screws directly and modeled and installed those lead screws.
The rotor diameter that I had settled on really didn’t feel to me like it would resist the deflections that I was trying to avoid despite the many bearings constraining it. I was worried about the large bending moments occurring at the backplate-spindle interface which could potentially cause chatter or weld failure.
I ended up doing a small amount of research into some standard spindle mounts, taking a liking to the American L0 taper. I knew that I could not realistically procure bearings and a rotor of the correct diameter to create such a massive nose-taper, so I formed my own using similar concepts and the same taper angle, but without the hollowed center characteristic of L0 tapers. While my first adjustment was to bring the spindle’s diameter from 0.75 inches to one inch, this dimension change did not last long, and I changed it again to 1.75 inches to support the tapered end that would allow backplates to be mounted concentrically on the spindle.
I specifically purchased a set of 1:2.5 bevel gears intended for repairing an angle grinder to use in my bevel system between the saddle-driving wheel and the lead screw. I used their dimensions and features to model them in Creo, then I formed an assembly comprised of a frame, wheel, the gears, bearings and the timing pulley to attach to my lathe model. I also decided at that point to switch out my aluminum flat stock, which I had intended to use as the saddle and cross-axis framing, with steel components for their superior rigidity and so I could weld structural attachments or choose a new thickness.
For the purpose of visualizing what it would look like and calculating belt length, I modeled and inserted the belt that would ride between the motor and spindle pulleys into all five of its possible locations. As I went along modeling these components, I began purchasing some of them, even if I thought that it was possible that I might use different components down the line. I needed some of them in my hands to know exactly how to model them and whether or not they would work for my purposes, and I could only find most of them online, preventing me from first checking them out in stores.
I originally wanted to weld a single backplate to the end of the rotor to simplify things but decided to scrap that idea for the tapered connection which required a retention ring clamping any chuck and backplate to the taper evenly, forcing good contact between the tapers. I modeled that retention ring system with course threads on both the ring and the generic backplate to pull the taper tightly into place, as it does in standardized tapers. When I later modeled my tail stock shaft and the components attached to it, I added a standard taper feature with which to attach a variety of centers including chucks, fixed and live centers, taps and boring bars.
Following my steel-focused changes, I decided that the hassle of casting the entire lathe bed wasn’t worth my time, and I replaced it with a similar weldable-steel bed. It was simple, formed from three long steel sections, the top face made from a one-inch-thick steel sheet with two half-inch thick, two-inch tall flanges on the sides to reduce bending and torsion of the bed. Having switched to steel sections from aluminum, I was no longer largely worried about the large moments caused by cutting forces on the saddle, so I further widened the saddle and its linear rails from 40 to 50 centimeters long, an increase of five centimeters in each direction.
I feared that the cutting forces would cause the elbow between the lathe bed and the motor housing to flex and cause chattering or inaccurate features with the setup that I had, having bolted them directly together between their contacting end surfaces. To increase the area footprint of the connection and prevent much of that potential moment, I added two triangular, vertically protruding one-inch-thick steel flanges, one to be welded to each side of the lathe bed, that would be bolted to the top of the motor housing and resist bending at the elbow joint.
As I mentioned before, I was unsatisfied with my idea to use a mill vise as a tool post, especially since it provided no way to attach a swiveling axis which would allow me to cut tapers. To remedy this, I removed the vice but left its base, then I modeled a new dovetail-based compound motion axis attachment to sit within the base, driven by an acme lead screw, on which I’d be able to mount a real tool post. I threw together a design for a generic tool post with reasonable dimensions and installed it onto the new compound, otherwise known as a top slide, and ensured that it would line up well with the axis of the spindle which is necessary for proper operation of the lathe. This way I could swivel my third axis to any angle relative to the spindle but in plane with the first and second motion stages, allowing me to cut tapers and use proper methods to cut threads without chatter.
Since I was no longer limited by the dimensions of my aluminum sheets and could find cheap steel scrap that would fit my needs, I widened the saddle a bit more, providing it with better resistance to bending moments driving it away from the spindle such as the axial components of cutting forces. I am aware of the fact that cast iron would have provided a superior lathe frame to the steel sections that I designed, but I have no large-scale iron-casting setup, nor do I have any scrap cast-iron to melt down, and welding a frame together from prefabricated sections is far easier than casting such large components in any home shop.
Cast iron is a porous material and provides damping properties to its structure, thus diminishing the noise produced by the system as well as modifying its resonant properties which could lead to defects in parts or a poor surface-finish. If I were capable of casting the components, much less transporting a lathe made from cast-iron, and I can barely manage to transport it in its current form, I would have gone with the cast-iron option. The benefits don’t outweigh the problems that it would cause in this case.
To provide a mounting mechanism atop the cross-slide for adding peripherals like vises, my top slide, and my tool-post, I extruded t-slots in its top surface. One of my main concerns with the linear rails that I ended up being rightly worried about was that the aluminum extruded sections on which they were mounted would not be straight due to their extrusion manufacturing process and wouldn’t be able to support the loads that I will be subjecting them to, especially those pulling the rails up away from the bed.
In favor of a far more rigid system with more spline points through which to control the aim of the rail and force it to be straight, I removed the aluminum supports entirely in my model, as I had also done with the rails that I had purchased. When I looked down their lengths, I saw complex curvature and became worried that the dimensions of the supports may not be consistent. Upon taking each rail apart to get a better look at how they were held together, I realized that only eight m5 machine screws were spread across the 1.2-meter-long rail pulling it flush with its aluminum extruded support.
To provide a new surface for the rails to be attached to, I decided that I would weld steel strips with a square cross-section atop the lathe bed and mill them flat afterwards. I would then hand-scrape the rail supports after they had been milled to bring them to the best possible flatness tolerances that I could achieve with my A-grade surface plate, ensuring with my spirit level and test indicators that the surfaces remained parallel after my fine-tuning with no twisting. Once these rail mounts were flat, I planned to mount the linear rails to them using 17 evenly-spaced m5 screws, ensuring that the motion plane of the saddle was as flat as the surfaces beneath the rails.
The trickier part was to come, requiring that the rails had no curvature end-to-end out of that plane of motion, and remained as parallel and straight as possible. My solution to that was to use set screws to drive small changes in the spline defining the rail’s curvature by pushing the mounting m5 screws from either side, forcing the spline to be straight through repeated checking against my surface plate using test indicators and fine adjustments of the set screws.
I realized at some point that the thicknesses of some of the plates of steel used in the model was excessive and that the weight that they added was wasteful for the small benefit that greater thickness added. I instead modified them to be thinner, changing the base, saddle and tail stock plate sections from one inch thick to half of an inch thick thus massively reducing the price and weight of the steel. To compensate for my reduction of structural rigidity, and to exercise some of the structural techniques that I had learned in my engineering classes, I added new protruding structural flanges to each plate, giving them bending and torsional resistance in every important direction.
On the tail stock, I added vertically protruding flanges on the front and back ends of the baseplate to resist the torsional moment deflection that would be inflicted between the tail stock tool and the bearings. To prevent deflection in the saddle’s overhang and in the cross-slide plate, I added two more braces to each. Using Creo’s analysis tools, I compared the polar, x, and y moments of inertia of the saddle and cross-slide structures’ cross sections and found that my flanges had increased the polar moments of inertia by factors of at least two, and most importantly increased the y moment of inertia by more than a factor of 10. This was the first moment when I truly realized how important the distributions of mass were in the rigidity of a structure and began to consider their benefits in every system that I designed from then on.
In my original Creo assembly containing all the lathe parts, I had never properly analyzed the relative locations of my linear rails and lead screws for optimization of the stages’ ranges of motion. With some simple math it was easy to maximize this length in both axes of travel and adjust the positions of the lead screw assembly and linear rails on the surface of the lathe bed.
Once I had modeled and added every structural component that I could imagine to the lathe assembly, I went back to each location that required fasteners and considered which fasteners would be sufficient, how many to add, and how to distribute them in their final locations. The vast majority of the fasteners that I selected ended up being either m5 or m6 machine screws with a few odd imperial nuts and bolts here and there. The lathe in its final state utilizes 196 m5, 133 m6, 5 m3, and 121 different imperial standard nuts and bolts and their many tapped holes, which hold everything together non-permanently allowing for maintenance and fine-tuning.
With the lathe in this nearly complete state, I began searching for the remaining components purchase and form the lathe from, with specific interest in acquiring old, used scrap steel from junk yards to repurpose and save on metal costs. When I went to the scrap yard called Marion Iron located in Marion, Iowa, I found much of the structural steel that I needed for the low price of 20 cents per pound. After much searching I was able to match the steel that I needed to build my lathe sections with moderate-condition steel stock that I’d found around the yard, which I then brought home.
Much of the steel was imperfect and required for me to rework my model. For example, I desired to form the bed of the lathe from three sections: A large flat section around 10 inches wide and a half-inch thick and two half-inch flanges, which required that I find a section large enough to cut them from. I was unable to find parts that met my exact needs which I could weld together to create this part, but I did find a large section of HSS (hollow structural steel) which I could replace the original modeled bed with, providing the model with far-superior bending and torsion properties without the addition of too much weight. With the new section having an external height of four inches and a width of 12 inches, I was able to further widen the bed in my model slightly, to slightly over 10 inches between the linear rails, limited to the top space between the corner rounds of the section which have a radius of about one inch.
Finally, after acquiring steel that I knew would work for the construction of my top slide, I was able to make modifications to its model, bringing it closer to what I had originally imagined. With these new changes and sections added to the top slide, I was able to add a wheel directly driving a lead screw to control its motion and remodel its connection points to the swivel-base supporting it to hold the base more flush with the top slide to provide better load resistance.
My newest changes to the lathe model occurred rather recently, after I began spending much of my free time on other projects including this site, being added sporadically as they came to mind. Having been dissatisfied with the bevel gear system that I implemented to drive the saddle through a wheel and lead screw, both because of its low load-handling capabilities and the amount of machining work that it would take to fabricate the assembly, I opted instead to purchase a pre-existing bevel gearbox online. I found that it was ex