After cleaning up the parts their dimensions were checked and minor adjustments made so they would fit precisely into one another.
Initial assembly allowed us to mark the exact locations for the fixing screws, which were then carefully drilled through the temporarily joint parts.
The lower end of the arm attaches directly to the “wrist” motor, making it effectively a key element of its structure. These holes were also carefully marked and drilled.
The arm was then temporarily clamped to a dummy shaft to access its overall stiffness.
Finally the arm was clamped to the aluminum shaft to test its integration with the robot.
Our tests so far demonstrated that the arm requisites are being met. However the aluminum shaft on which it is clamped has proved to be too flexible causing an unacceptable degree of torsion to the whole assembly.
The voind robot design is based on a swivel arm that rotates on a horizontal shaft driven by a worm/wheel combination gear. This shaft has a very simple design with a 10mm main section and a 20 mm narrower section with 6mm diameter.
To fabricate the narrower section we decided to build a small combined Milling/Lathe machine from MDF board. This contraption basically holds firmly a small rotating tool with a small drill and at the same time allows a 10mm shaft to be fed and rotated against the mill.
After a quick test we were good to go at machining the end of our aluminum shaft. Check the actual footage below:
After the operation the machined surface shows a scaled aspect due to the manual feeding of the shaft.
But after a quick sanding with fine sand paper the result looks very smooth:
You can see below the wheel firmly clamped to the new shaft and being driven by the worm.
This aluminum shaft proved to be too much flexible. A steel shaft will have to be used instead.
One of the motors on the voind robot is an External Linear Stepper Motor, more precisely an SY17STH0404-300A6.35X2X2-P1-EX. It has a threaded shaft engaging a threaded nut that drives the robot arm perpendicularly to the equipment face.
This motor is conveniently supplied with a 30 cm shaft which should be more than enough for every application. However in our case we needed to cut the shaft shorter.
We first wrapped the motor in plastic film to protect it from any metal chips and used a small drill press vice to hold it firmly by the shaft, having wrapped it in a generous quantity of soft paper to protect the thread.
We then used a Dremel EZ456 cutting wheel to make an initial cut one centimeter outside the mark to be able to position the tool perpendicularly to the shaft for the final cut. To prevent overheating the shaft and motor we cut in 10 to 15 seconds intervals followed by 45 to 60 seconds periods of pause for cooling.
Interestingly, although the shaft seems to be made from stainless steel, not a single spark was produced during the cut…
Ideally we would have a 6.35×2 nut screwed in to help restore any damaged thread start by removing it after the cut. However we did not have any suitable nut around and we had to resort to a magnifying glass and some fine sand paper to make sure the threads started smoothly.
Finally the motor with the correct thread length and ready for assembly.
Some of the parts for the voind robot project are built from bent sheet metal. This is the case for the two stainless steel chassis side plates on which the main ballscrew is mounted:
Fortunately, having sourced the chassis construction to a very competent expert on metal fabrication, I never had to account for the material along the bending lines and only specified the final dimensions of the finished parts:
The failure to understand how to correctly incorporate the bending into the flat part design was partially responsible for multiple attempts at this part fabrication:
We finally got it right after using a mockup of the part which included the main bends but using only a narrow cross section of the finished part (the part on the upper right image above). Taking precise notes of the bending lines and adjusting them accordingly on the production part landed us with a precisely dimensioned part (almost) ready for integration!
More bending action lies ahead as we prepare to fabricate the more complex Y Assembly Cable Guide supports.
For the voind robot project we are using nearly two dozen screw references, mixing types (DIN 7984, 912) and sizes (M2 to M5, 6 to 35 millimeters lengths).
The screw types and positions were established during the design phase taking into consideration the dimensions of the parts, the forces involved, the clearances for the screw heads (and nuts) and also how to reach the screws during assembly and maintenance.
Most of times the screws had mating threads cut into one of the parts using a manual tap.
Some tap operations proved to be very challenging, specially when cutting M3 threads into stainless steel beams and in particular the one where we could not use a tap handle due to the part geometry:
In the end we assembled quite an impressive array of screws, nuts and washers references!
This assembly supports the main cable guides carrying power and data for the X Assembly carriage and also provides a mounting point for the micro switches used to detect the Y movement limits.
This part is being built from aluminum and acrylic sheet.
Several arrangements for the micro switches are still being evaluated.
One technique we are using here to save precious space and minimize exposed metal near electric contacts is embedding the nuts into the acrylic using a small soldering iron to heat the nut and press it into the acrylic part.
The final design for the aluminum sheet part is not yet decided. Until then we keep fabricating several versions of it 🙂
This was the first time we ventured into aluminum fabrication using a CNC milling machine. The majority of the voind robot parts were designed to be fabricated from aluminum bar using a CNC milling machine.
Soon into the project we realized how expensive it would be to have these parts sourced from specialized shops. We were quoted with hundreds of euros for a single aluminum part and we would need to fabricate more than 20 different parts!
Online metal fabrication shops like eMachineShop were also evaluated but were still too far outside our budget.
Apart from this we feared we would end up doing several iterations before we obtained exactly the required part, specially because we lacked the knowledge and experience to spec the parts correctly.
So we decided to try and do it ourselves all the way from the 3d files up to the actual part machining.
We started by visiting a local fablab shop where we met knowledgeable people who introduced us to the basic tools and techniques used to create the parts drawing files and prepare them to be submitted to the milling machine.
Unfortunately this particular fablab shop had their milling machine already booked for the dates we planned. However they kindly introduced us to another fablab shop where we could book the milling machine for specific dates which were compatible with our calendar.
Isn’t it wonderful to have access to such an array of resources, tools and knowledge?
If you happen to have a project or idea and you are stuck with how to realize the physical aspect of it, search for the closest fablab and pay them a visit.
Mills and Milling parameters
After several broken mills we set on 3mm HSS + 8% CO end mills:
Spindle: 5500 rpm
Feed rate: 5mm/s
You can buy these mills from your local specialist tool shop. I bought mine here and here.
Lubrication and cooling
During the milling operation we manually sprayed cutting lubricant and also water to cool things down, specially during the most aggressive phases of the cut, for example when the tool had to start a new cut into the material.
Sometimes we could feel the generated heat just by touching the aluminum up to 15 cm from the cutting action. The only time we tried to save on the lubricant and water we ended up with a mill literally welded to the aluminum. Not wanting to ruin another mill we diligently kept the spraying action throughout the whole operation.
We used MDF as sacrifice boards on which we fixed the aluminum bars using screws. The exact positions of the fixing screws were determined during the layout of the files in preparation for the milling operations. This way we could plan for an affective support of the material during the mill operation and at the same time check that the tool path would not touch any of the screws.
The MDF board was then securely fixed to the milling machine bed using additional MDF scrap boards and clamps.
For long sequences of operations where the MDF board would be exposed for significant time to water and lubricant, we first covered the board and edges with adhesive tape.
However, any drilling operation that drilled into the MDF board acted as a direct injection point of water into the MDF causing it to swell and slightly push the aluminum up, enough to ruin all accuracy in the Z dimension. This led to some ruined parts and on two occasions left parts dangerously loose due to overshooting the cuts in the z axle.
The software we used to create the toolpaths did not offer enough flexibility to let us specify where to place brigdes to hold the cut parts in place. As a result we decided to leave a half millimeter skin holding the parts and separated them afterwards.
Workflow and Planning
The richness and openness of the environment at a fabrication shop like a fablab can quickly draw your focus away from your main objectives. There will be plenty of experiences to share and projects to admire and, unless you prepare yourself with a robust work plan, you might end up not using your time and available tools in the most efficient way.
Before a visit to the fabrication shop I usually try to:
have a detailed plan for the tasks I want to execute;
mentally rehearse and visualize all the tasks, the tools needed and all the contingencies which are likely to happen (like a broken mill, unavailable PCs, a corrupted file, etc.);
prepare a checklist on the day before and make sure I am carrying everything I might need;
Most of the times the freshly cut parts needed extensive finishing before being ready for assembly, specially because we insisted on leaving a half-millimeter thick “skin” at the bottom of the cuts to hold the parts together.
For that we used a Dremel and a mill tool to trim away the skin with the assistance of a steel guide to avoid cutting into the part. A final touch with a file and sand paper gave the part an acceptable finish to move it to the test assembly phases.
Fixing parts together
We looked for some time into the possibility of welding the aluminum parts. However due to design constraints which dictated that some assemblies would have to be disassembled for maintenance, we drill and tapped holes and used mostly DIN 7984 M4 screws to fix everything together.
Only time will tell if the joints are strong enough to withstand vibrations during normal operation.
The voind robot concept was developed virtually in great detail even before a single part was produced, enabling us to work out the best arrangements to achieve a compact footprint (9 cm in height) for the sliding carriage assemblies. This part of the design was done using the excellent open source 3d design and animation package blender.
Once the overall chassis dimensions were established we produced the technical drawings and sourced the chassis fabrication and welding – stainless steel and tig welding.
We chose to use a low profile linear guide system from igus, allowing us to fit the entire X-Y carriages and motors within a 90 mm height.
For the aluminum parts we produced the 2.5d drawings using DraftSight
These drawings were then combined and laid down for cutting and machining from stock aluminum bars.
First attempts at cutting aluminum were disastrous!
After some tests we settled on conservative settings and HSS/CO 3mm cylindrical end mills:
Spindle: 5500 rpm
Feed rate: 5mm/s
A sequence of machining sessions followed – totaling around 20 hours – during which the majority of aluminum parts were cut.
We tried to keep the maximum possible dimensional accuracy with the tools at our disposal. This meant that sometimes we would be positioning a part for over 10 minutes only to perform a single drilling operation.
As soon as the first aluminum parts were fabricated we proceeded into initial tests of the assembly fits. These tests validated the dimensional correctness of the parts and the overall clearances which were calculated during the early stages of the design.
For example the clearance between the ballscrew and the surface of one of the parts is less than one millimeter across the entire width of the robot. This is one of several design decisions which allowed us to cram three DOFs (degrees of freedom) with motors and associated mechanics within a 9 cm height footprint.
The control electronics is being built and tested side-by-side with the mechanical components. In due time it will be attached to the robot chassis but until then it is laid out on an area of the assembly zone protected by a plastic transparent cover.