Announcements OpenRUGDMMC

Public domaining of ideas

posted May 24, 2019, 7:18 AM by Anthony Douglas

I got a bit worried over the last couple of days that there might be patent encumberances on some of the ideas here, because it has become clear that there are some key ideas that must not be encumbered.

The fundamental idea of rough deposition followed by milling is under control because it is already published as mold sdm.

However there is one other idea, which is the use of ultrasonic consolidation of powders as a deposition method.  This is very important.  Although it is not terribly useful for making objects directly, it has a great deal of promises for making molds.  

So, I am writing this post now expressly for the purpose of getting this idea out in the public domain so that it cannot be patented.  

The idea is to use the application of pressure and ultrasound, with any mode of vibration including transverse or longitudinal, especially relatively low pressures (in order to give a reasonably practical machine), to consolidate powdered materials such as:

Water soluble materials like salt, urea, or other chemicals.

Hexamine, hard waxes and other materials which are only soluble in substances other than water.


-soapstone or other soft materials which are not water soluble.  These may bind even without ultrasound.

-natural graphite, or synthetic graphite with additives such as niobium or manganese or iron, or other transition elements or additives.  These may be added by ball milling the graphite with finely powdered additive material, or the dissolution of some of the metal in a solvent, mixing it with graphite powder and evaporating the solvent, then maybe milling the result.  This can produce a "sticky graphite" which has almost all the same properties as graphite with regard to machineability, softness, and yet being highly temperature resistant (which allows it to be used to cast titanium, steel and other materials.) , but will stick to itsself much better, and can be compacted into a solid mass with significant strength.  Some patents indicate that mixing pure synthetic graphite with 1 percent manganese and compressing under 2000 psi give a solid with compressive strength of 500 psi.



There are documents describing the use of this in pharmaceutical tablet manufacture, and there is a couple documents in which people are describing the consolidation of metal powders, but rarely for additive manufacturing.  There is only one document that I have found that describes investigation of such ultrasonic consolidation of powders for additive manufacturing, and that is by the company that does the Ultrasonic Consolidation 3d printing like system, which uses normally foils or tapes, but they investigated using aluminum powder, too, just with computer simulation, which isn't really that valuable.

We need to do some experimenting with this kind of material deposition approach asap.  That and the process planning are the only key barriers.  Once you have that, you just add an auto change routing machine and aparatus to do the deposition, and basically the gcode can be run to produce a mold immediately.



work continues, very slowly

posted May 16, 2019, 4:17 PM by Anthony Douglas

Things are continuing.  My current focus is to try to create a plugin in fusion 360 which will allow the automation or semi automation of the CAM programming process to the extent that the use of very small layers is possible.

This is critical, because although at this stage a basic test part could be produced, it is extremely time consuming to program the layers.  The current version of the instructions I have for the process can be found below.  You can see how labor intensive it is.

Ideally, you want to use layers in the range of a millimeter or half a millimeter high, which means a lot of layers.  This allows you to employ rest milling with very small diameter tools to get those fine details and features, small radius corners etc. without resorting to five axis milling.  This is one of the main strengths of rugdmmac, so it is important to realize this.

I have been thinking a lot about how to do this, and I think if I can use python to do a sort of test;  the model must be divided into what I would dub major layers, the layers of maximal thickness that could concievably be milled with any end mill, which could be as long and thin as you would like.  These major layers will then be subdivided into smaller layers of equal height that are less than a millimeter or whatever thick.

Basically the model must be decomposed into machinable "compacts".  These are solids which can be machined.  That is, the tool can reach in there.  In reality it may require a tool that is too long and thin to actually machine perfectly, but for a moment we can ignore that.  That will come naturally as we make the layers thinner.

The problem is that we cannot simply use say 1 mm layer height and dumbly put the layer tops wherever we want without eventually getting defects in the part in various ways.  The thing is that at the so called draft transition areas, you get overhangs, and the material under the overhang cannot be machined.  However, any part can be decomposed into segments that have no overhangs.  This is a key realization!  Things like this reality are what make things possible.

Thus, we can ultimately make any part of any shape, it's just a matter of making the layers begin and end at the right z heights.

I believe I may be able to do this by first making a test which determines if a layer is makeable: Suppose for any candidate solid, that is, a slice of the model, you create an array of closely spaced lines, spaced say a hundred microns apart.  Rays, we can call them.  They radiate from the bottom up.  They will either never encounter the solid, or they will pass through the solid and out the other side (since there is no solid above the top of the slice).  However if they pass into the solid and then out and into it again, that layer is invalid.  It means the z height must be reduced until all the rays either do not enter the solid or enter and then exit only once.

I think this should work.  

Once we have a test like this there are various ways we can use it to analyse the model.  You could do a binary search of all possible z heights, starting from zero, to determine the highest z height that gives a solid which passes the test.  Or you could draw rays all the way to the top of the model, consider all invalid rays, and take the shortest invalid ray (invalid meaning it enters the solid, then, impermissibly, enters again (which implies it must have exited at some point)) to be your layer height.  

You could then further refine the layer height, but for now I think we can put up with any slight innacuracies this leads to.

Then, you would record that z height as the first layer top, and then cut away the model below that z height, and repeat the process, until you have a set of z heights for the layer tops.  

Then, take the model, surround it with the CAC material in the modelling software, create planes or surfaces at the relevant heights, slice both solids (the cac block and the model) at the relevant heights, then we can use those surfaces to do the cam.  That part also has to be semi automatic, and there are additional challenges that are faced there.

So essentially every layer except possibly the first and last, will contain both upward facing surfaces and downward facing surfaces.  First the blue is deposited, then machined.  Then red is deposited, then machined, then cleared away anywhere blue needs to be deposited.  Then blue is deposited, and machined, and this will gouge the red solid but not affect the finished surfaces. Then red is again deposited, and machined.  The process repeats.  The machining process is basically the same, just a pattern of operations and tools with rest machining enabled.  It can be the same pattern.  The surfaces will need to be selected accordingly.

work continues, very slowly

posted May 16, 2019, 4:16 PM by Anthony Douglas

Things are continuing.  My current focus is to try to create a plugin in fusion 360 which will allow the automation or semi automation of the CAM programming process to the extent that the use of very small layers is possible.

This is critical, because although at this stage a basic test part could be produced, it is extremely time consuming to program the layers.  The current version of the instructions I have for the process can be found below.  You can see how labor intensive it is.

Ideally, you want to use layers in the range of a millimeter or half a millimeter high, which means a lot of layers.  This allows you to employ rest milling with very small diameter tools to get those fine details and features, small radius corners etc. without resorting to five axis milling.  This is one of the main strengths of rugdmmac, so it is important to realize this.

I have been thinking a lot about how to do this, and I think if I can use python to do a sort of test;  the model must be divided into what I would dub major layers, the layers of maximal thickness that could concievably be milled with any end mill, which could be as long and thin as you would like.  These major layers will then be subdivided into smaller layers of equal height that are less than a millimeter or whatever thick.

Basically the model must be decomposed into machinable "compacts".  These are solids which can be machined.  That is, the tool can reach in there.  In reality it may require a tool that is too long and thin to actually machine perfectly, but for a moment we can ignore that.  That will come naturally as we make the layers thinner.

The problem is that we cannot simply use say 1 mm layer height and dumbly put the layer tops wherever we want without eventually getting defects in the part in various ways.  The thing is that at the so called draft transition areas, you get overhangs, and the material under the overhang cannot be machined.  However, any part can be decomposed into segments that have no overhangs.  This is a key realization!  Things like this reality are what make things possible.

Thus, we can ultimately make any part of any shape, it's just a matter of making the layers begin and end at the right z heights.

I believe I may be able to do this by first making a test which determines if a layer is makeable: Suppose for any candidate solid, that is, a slice of the model, you create an array of closely spaced lines, spaced say a hundred microns apart.  Rays, we can call them.  They radiate from the bottom up.  They will either never encounter the solid, or they will pass through the solid and out the other side (since there is no solid above the top of the slice).  However if they pass into the solid and then out and into it again, that layer is invalid.  It means the z height must be reduced until all the rays either do not enter the solid or enter and then exit only once.

I think this should work.  

Once we have a test like this there are various ways we can use it to analyse the model.  You could do a binary search of all possible z heights, starting from zero, to determine the highest z height that gives a solid which passes the test.  Or you could draw rays all the way to the top of the model, consider all invalid rays, and take the shortest invalid ray (invalid meaning it enters the solid, then, impermissibly, enters again (which implies it must have exited at some point)) to be your layer height.  

You could then further refine the layer height, but for now I think we can put up with any slight innacuracies this leads to.

Then, you would record that z height as the first layer top, and then cut away the model below that z height, and repeat the process, until you have a set of z heights for the layer tops.  

Then, take the model, surround it with the CAC material in the modelling software, create planes or surfaces at the relevant heights, slice both solids (the cac block and the model) at the relevant heights, then we can use those surfaces to do the cam.  That part also has to be semi automatic, and there are additional challenges that are faced there.

So essentially every layer except possibly the first and last, will contain both upward facing surfaces and downward facing surfaces.  First the blue is deposited, then machined.  Then red is deposited, then machined, then cleared away anywhere blue needs to be deposited.  Then blue is deposited, and machined, and this will gouge the red solid but not affect the finished surfaces. Then red is again deposited, and machined.  The process repeats.  The machining process is basically the same, just a pattern of operations and tools with rest machining enabled.  It can be the same pattern.  The surfaces will need to be selected accordingly.

work continues, very slowly

posted May 16, 2019, 4:15 PM by Anthony Douglas

Things are continuing.  My current focus is to try to create a plugin in fusion 360 which will allow the automation or semi automation of the CAM programming process to the extent that the use of very small layers is possible.

This is critical, because although at this stage a basic test part could be produced, it is extremely time consuming to program the layers.  The current version of the instructions I have for the process can be found below.  You can see how labor intensive it is.

Ideally, you want to use layers in the range of a millimeter or half a millimeter high, which means a lot of layers.  This allows you to employ rest milling with very small diameter tools to get those fine details and features, small radius corners etc. without resorting to five axis milling.  This is one of the main strengths of rugdmmac, so it is important to realize this.

I have been thinking a lot about how to do this, and I think if I can use python to do a sort of test;  the model must be divided into what I would dub major layers, the layers of maximal thickness that could concievably be milled with any end mill, which could be as long and thin as you would like.  These major layers will then be subdivided into smaller layers of equal height that are less than a millimeter or whatever thick.

Basically the model must be decomposed into machinable "compacts".  These are solids which can be machined.  That is, the tool can reach in there.  In reality it may require a tool that is too long and thin to actually machine perfectly, but for a moment we can ignore that.  That will come naturally as we make the layers thinner.

The problem is that we cannot simply use say 1 mm layer height and dumbly put the layer tops wherever we want without eventually getting defects in the part in various ways.  The thing is that at the so called draft transition areas, you get overhangs, and the material under the overhang cannot be machined.  However, any part can be decomposed into segments that have no overhangs.  This is a key realization!  Things like this reality are what make things possible.

Thus, we can ultimately make any part of any shape, it's just a matter of making the layers begin and end at the right z heights.

I believe I may be able to do this by first making a test which determines if a layer is makeable: Suppose for any candidate solid, that is, a slice of the model, you create an array of closely spaced lines, spaced say a hundred microns apart.  Rays, we can call them.  They radiate from the bottom up.  They will either never encounter the solid, or they will pass through the solid and out the other side (since there is no solid above the top of the slice).  However if they pass into the solid and then out and into it again, that layer is invalid.  It means the z height must be reduced until all the rays either do not enter the solid or enter and then exit only once.

I think this should work.  

Once we have a test like this there are various ways we can use it to analyse the model.  You could do a binary search of all possible z heights, starting from zero, to determine the highest z height that gives a solid which passes the test.  Or you could draw rays all the way to the top of the model, consider all invalid rays, and take the shortest invalid ray (invalid meaning it enters the solid, then, impermissibly, enters again (which implies it must have exited at some point)) to be your layer height.  

You could then further refine the layer height, but for now I think we can put up with any slight innacuracies this leads to.

Then, you would record that z height as the first layer top, and then cut away the model below that z height, and repeat the process, until you have a set of z heights for the layer tops.  

Then, take the model, surround it with the CAC material in the modelling software, create planes or surfaces at the relevant heights, slice both solids (the cac block and the model) at the relevant heights, then we can use those surfaces to do the cam.  That part also has to be semi automatic, and there are additional challenges that are faced there.

So essentially every layer except possibly the first and last, will contain both upward facing surfaces and downward facing surfaces.  First the blue is deposited, then machined.  Then red is deposited, then machined, then cleared away anywhere blue needs to be deposited.  Then blue is deposited, and machined, and this will gouge the red solid but not affect the finished surfaces. Then red is again deposited, and machined.  The process repeats.  The machining process is basically the same, just a pattern of operations and tools with rest machining enabled.  It can be the same pattern.  The surfaces will need to be selected accordingly.

Progress report

posted Apr 9, 2019, 7:44 AM by Anthony Douglas   [ updated Apr 9, 2019, 10:39 AM ]

I have made some substantial progress in the last month and a half, mostly because I have had a vague semblance of a workspace, a section of a shared two car garage, made possible by my friend and fellow open source ecologist Matt and his wife Carolyn.

I have been able to make a lot of progress on figuring out the core of how the process works, the process for generating the gcode, and formulate plans for what material to use at first, and the deposition apparatus.

However I am still working in what are objectively extremely poor conditions, and this slows everything down severely.  My housing situation is again precarious, too. 

As long as I continue to be allowed to use the garage, I can still progress at a slow pace at least.  I am discussing cleaning up and reorganizing the garage etc. and may be able to get better housing soon.

I have made a few videos that explain the process, as it evolves things change fast though, so I will only bother to post the most recent one here.

I am currently focussing on producing an add on written in Python, for fusion three sixty, which I think will be a practical way to automate the process to produce gcode.

I have achieved the very important first milestone of producing the complete gcode for a test part, with a great deal of manual interventions and thick layers.  The layers must be thick to limit the number, or the process is far too time consuming.  As it stands, with the use of thick layers, the feature size is limited though.  Much of the power of the process comes from being able to use small layer heights, thereby allowing the use of small diameter tools to get those small features machined.  

With the manual approach, you get a substantial radius on the inner, upward facing corners.

So, automation of this process is critical, however it looks promising that a plugin will be able to do this.  

VIDEO REPORT: https://drive.google.com/open?id=1C-93TU5d1I3L7_FnowKFWhvPsdmvIUDO

current instructions for manual cam:

prepare the model
-import the model into solidworks.  create split line
-offset all upwards facing surfaces, have to manually select them. careful not to miss any, sometimes it splits faces in nonsensical places
-offset all lower facing surfaces
- offset all vertical surfaces.
-import the solidworks file into fusion.
in fusion from now on:
-have the solid body present. 
-put all up surfaces in a folder
-put all down surfaces in a folder
-put all vertical surfaces in a folder.
-color all up surfaces red
-color all down surfaces blue
-color all vertical surfaces green

**Start layer height decision making, splitting surfaces and sketching

- draw a sketch under the model to indicate the deposit layer boundaries. this will be used for extruding stock blocks. call this the deposit layer boundary.
#we could make a 3d sketch with points for layer heights.  little point for now. in the future allows easier editing ands positioning of planes, esp where the self occlusion point are which is hard to position a construction plane at.
- draw another sketch offset 1/2 of a tool diameter outwards, to use as the machiing boundary when doing red pocketing with hard exclusion boundaries. So 1/16 of an inch out if we are using 1/8 inch end mills for roughing.
-decide on a compact top height based on: 


-4 major hard rules, which determine major layer heights, which makes layers regardless of tool aspect ratio limits, to ensure machining is possible with a three d strategy. violating these results in defects, however we cannot use thin enough planes for all features yet, so that criteria we have to slide on till we have a plugin:
-- blue top cant be raised above the point where it occludes a red surface.
-- a red top cannot be raised above where red occludes a blue
--no top can be raised above a point of self occlusion, even a red top above a blue self occlusion point. I think.
-- the max layer height that is practical for the tools used and feature sizes or sharpness.
Guidelines we can use for now to narrow down the choices:
 -- we could make the red top coincident with the blue deposit top in some cases. that would be consistent with the minimal red advance strategy.  indeed it requires it.  that would simplify things in some ways. it could prevent sharp vertical edges and some sharp points of red which might break off and result in a defect. ideally we would specify the red solid shape and then machine it, not the tool layer boundary stuff but gotta make do for now. we camn do the contour first, that would revenet sharp points.  sharp points are over blue anyways so they will get redone so doesnt mater.  dpesm't jhelp with the vertical sharp surfaces. the tool does not need to go over ther though maybe it wont in which case problem solved.doesnt really help much once you have the other stages automated
-- red should be minimally above the blue.  that implies raising blue as much as you can get away with, generally.

-make a surface for the plane, might as well make the sketches for them projected rectangles of the deposit layer boundary.
-Make the deposit top surface for each plane using the surface offset tool
- color it. we only ned to color the top of each pair really.
-the program should name them, when doing it manually lets not bother for now. we can see thm visually easily enough.
-split all surfaces that intersect with the deposit top using the deposit top. #lets try splitting surfaces only at the deposit top. not compact top Open split face tool, first selecty the splitting surface, then view the model side on and use a crossing box so drag from the top right down and left across the model at the height of the layer to select surfaces that intersect with the layer. for surfaces below it, we can select the whole thing all the way down, and machining will still stop at the stock bottom without issue.  

-for both blue and red layers, make a sketch with the intersection of the compact top planes and solid model. slice the sketch and select the model. then select the patch to do the next step without exiting the sketch so it stays sliced. I think. we might not always need it.  for now lets make it just in case we need it later on. for the first blue layer we needa special sketch  cuz the blue is not at the level of the red bottom.

- for red make a patch of this compact top intersection sketch created, this is useful as a machining surface for blue first layers to expose the red again. 

-- for red layers, then make a sketch on the red deposit top.  project in the inner part of the red layer tops, which would be the sharp upper edges if just the red was machined. or all of them. project in and combine it with the lower blue compact top layer intersection sketch.  identify the tool exclusion contour. close it with one line. that makes the inner machinig boundary, to keep the tool out of.   when the red does not intersect the olane there would be only one contour, and we would just have no exclusion contour. the program has to account for this, not go and treat that contour as a tool exclusion zone.  Or just say that in the event the red surfaces do not intersect, dont bother making a sketch or selecting the contour. the machining operation generation section has to also account for this. #this region doeslt reallyneed to be closed witha straight line, could use the section of the compact top intersection sketch to close it.  this is a bit simpler, we just say if there is more than one countour, its the innermost one that is the tool exclusion zone.  YEs, this also prevents any lumps of blue getting in there, which could actually happen otherwise.

-return to "-decide on a compact top height" and repeat for the next layer, remember minimal red advance, maximal blue

**make stock blocks
- next, make the stock blocks of red and blue material to use in hsmworks. use the sketch of the layer boundaries previously produced, and a two direction extrusion, one direction to the top one direction tothe bottom of the layer. 
-for blue make it extend from compact top of lower layer tp deposit topof iupper layer
-fOR RED make it to the compact top of the previous red layer. The machining heights are what we are selecting here. lets make it extends from deposit top to compact top. 
-make a new folder for the red stock and name it red stock
-make a new folder for the blue stock and name it blue stock
-put the relevant stocks in the relevant folders so you can find them when camming. divide and conquer.
- put the little used surface bodies in a littel used surface bodies folder so they dont slow you down.
# we can use the volume of the stock blocks to estimate the maximum amount of material needed to deposit. for now just let the extra go to waste. not right now


**in hsmworks:
dont forget to use the radio buttons for selection 
for the first two layers, make new setups. then, for subsequent layers, duplicate and edit those. Then after the duplication, delete some of the operations, change the model geometry in the similar one, duplicate them again, then change the tools. This is just to avoid reselecting the geometry. or we could just duplicate and reselect geometry for all operations sure 

there are three things to do in each layer: clear the way for the next deposition, i.e. no material taking up volume which is going to be consumed by the next deposit with different material.  two, expose the top of material so the next deposit sticks.  three, clear material away from the surfaces to be shaped in that layer.
-for each new layer:

-Seed setup: blue:one pocket operation, none selected for model geometry for both setup and the pocket operation, let use machine shallow stepovers, minor preference.  
red:for red, three operations, one pocket for the top area with boundary the red surface projection combined with compact top sketch and the layer boundary, a 2d contour to trim away red with the contour being the compact top sketch of the blue layer below, and another pocket, with the tool exclusion boundary the compact top intersection sketch of the blue compact top below, and the offset of the layer stock boundary.
-- we can make the top the stock top and bottom the model bottom for the blue operations, having made the stock blocks accordingly.  The red needs different heights for the different pocket operations.
-- make the entry plunge
-- stock to leave, 10 microns to make the simulations look better. 
- max step down 3 mm but with machine shallow areas so it basically does some finishing but comput time is less
-duplicate and reposition it
-modify the setup for the new stock block.  make sure there is none selected for the model.
-then modify the furst pocket operation for model surfaces and machining boundaries. dont forget to select the surfes in the layer below, which are inthe overlap region
--for red first layers, for the first pocket operation, make the machining boundary the area between the two or more contours that we made as sketches for the purpose previously, the projection of the red surfaces combined with the model and compact top intersection sketch to find the smallest inner contours made from the lines that close the contours where the two sketches intersect.  Make the boundary tool center

-- for blue layers, dont forget to  select the patch on the top of the red layer below and machine it with pocket to clear way for red deposit and expose the red surface
- for red layers, do a contour operation to chop off extra red, after the first pocket.  the contour being the model and compact top intersection sketch in the blue layer below, on the compact top of the blue layer below.
- for red layers, do a second pocket operation to the bottom of the stock, with a tool exclusion contour that protects blue.  One contour will be the compact top intersection sketch, the other the sketch slightly larger than the stock boundary that I made way back there.

- maybe make more operations with smaller tools and rest machining and/or make derived operations for finishing. #at this point the number and variety of operations to try to acheive closer tolerances and better feature sharpness for internal corners by using smaller tools etc. is mostly discretionary. lets leave most of that till later.  just a matter of more tools, tighter tool paths, thinner layers.  That we need the plugin for because smaller tools take smaller layers and it is impractical to program a large number of layers.
-now repeat for the next layer.
 
- add the layer height and stock volume nc code to each layer.

Words from the project initiator

posted Dec 18, 2018, 1:20 PM by Tiberius Brastaviceanu

It will overcome some of the most important barriers to open source hardware production, resolving the absurd situation we find ourselves in, in which we cannot practically make even the most humble of the objects that form more complex hardware systems. Any mounting bracket made from metal, or object which is of larger size (such as a foot across, to the size of a vehicle chassis), or cannot have the poor tolerances and ridges that come with 3d printing, is a serious barrier.  When we encounter dozens of these, it scuttles prospective projects.

OpenRUGDMMAC will solve this.

I am currently trying to secure a room to work on it.  I have some $12k in funds saved up to give myself some time and money to work on it, but crowdfunding is a high priority.  

The immediate main stage, right now, is to produce a document describing the mark 1 prototype, and high level process algorithms in English or pseudo code.  I need to collaborate with others, in particular on the software for process planning, including a form of slicing at the right z heights based on model geometry, the generation and conversion of the machine agnostic path to a set of gcode programs, which have the deposition steps interleaved with the machining.  None of that is cutting edge.  However it would not be easy to produce from scratch.  The path planning for 3d rest milling is the hardest part.  There is extensive work done on this in the technological ecosystem which we can leverage to relatively rapidly get things set up.

In the very near term, to drum up collaborators and money, the production of a good set of seed documentation, including describing the process and the importance of it, which the long document is a start on (needing editing down and some animations more than anything), and to focus on what I would dub a mock up of the process, seems like a good approach.

For the mock up, I can draw a good test part in solidworks, then simply produce the machining gcode by slicing it manually, and using hsmworks to do the cam programming.  This produces the gcode required for all machining.  I am in the prices of trying to secure an auto tool change mill/router whose characteristics I know, having used it for some months previously.

The deposition step will require at least a modest apparatus.  However I may simply resort to spreading it on the layer like peanut butter, to keep everything very simple. The reality is that to do thing manually until it is more clear what works, is probably a good idea.  It is not clear how long it will take to set, if carbon dioxide may be used to speed setting, if the material can be dispensed from a syringe or piston or if some other means must be found, if vibration will be important, or whatever.

Supposing the layer height is a millimeter, and the build volume is five cm by five cm, a five cm block would take five hundred layers.  One thousand deposition steps, one for each material.  Clearly the layers must be rather thick, or it must be automated for more serious testing, then.  Also layer deposition time cannot be long. The layer height will often be much higher, if the material can set at such depth. The materials used in so called ceramic block casting can be applied as bulk solids.

The possibility of causing powder to bond together under ultrasonic ation and pressure is very interesting.  There is a document I found in which the authors were reporting successfully using 9000 pascals of pressure, 20 kilohertz at 1.8 microns amplitude vibrations in the direction parallel to the upper surface of the powder bed, to successfully bond 10 microns aluminum powder.  This is slightly surprising, as powder consolidation usually employs much higher powers and pressures, however they are probably tolerating lower densities of the resulting solid.  I wrote to the author.  They say they have more papers under review describing their experiments.

I have found a document describing the use of 1 percent by molarity of a transition metal powder of 325 mesh, combined with powdered graphite (mesh size I dont recall or was not listed) to grt the graphite to effectively bond together under pressure only.  Ultrasonication will probably enhance this.  The material had500 psi compressive strength, which should do for us.

Ultraso ic transducers in the 100 watt 20-40 khz tange are available through e.g. Amazon easily.  Higher powers should not be hard to obtain.  1.7 mhz is commonpy used in fog generators.  Some common machines I found use ten disks of a total of 25 watts each.  It remains a matter of hypothesis if higher frequencies would be helpful.  Also whether that is the input power or the output power, frankly. It appears practical to stack such transducers with channels for cooling water flow between them to achieve higher power intensity.  Unfortunately although promising, ultrasonic binding looks like it will take significant developme t effort, which may be unwise at this stage.  It may be more advisable to use binder and powder, such as low viscosity wax and a water soluble wax, instead.

Colloid makes a very important potential binder, silica or zirconia or even graphite colloids especislly..  it epuld be good in genersl for aluminum, plastics andgel cast ceramicsand useable for steel probably, with some precision loss.  However drying time may be an issue.  There are however many ways of causing colloids to gel, indeed the hard part is generally getting them to not gel.  We increase the colloidal concentration until it is easily destabilized by changes in ph or temperature, for instance, and then drop the temperature, or raise it upon deposition by ensuring the syringe containing the undeposited materisl was at a different tempaerature.  Or directing a straam of carbon dioxide at it to change the ph might work.  Gasses like ammonia are used industrially to do this during investment cadting coating operstions.  However they are slightly hazardous and harder to obtain.  Freezing works.  There are many options.  

The water in the colloidal system would dissolve any water soluble components in the support material.  There are colloidal systems based on alcohol instead of water, but it is probably more adviseable to use a system for the support material that is alcohol rather than water soluble instead, becausr yhe chemistry of water based colloids is much better known and manipulated.  A wax and an as yet unitentified solid, alcohol soluble material would be good.  Or soluble in hexane or some other solvent. Acetone is a bit harder to work with. They are all flammable. But we can roll with that for now. We need to peoduce test parts to bring in interest and money.  A common filler used in combination with wax in investment casting is cross linked polystryrene.  That might be a good choice.  The styrene monomer might do, too.  Really any random solid that is available in suitable form, low cost, recycleable, is the main thing, but I dont have many candidates right noe.  Perhaps a better approach is to start with a survey of easily available powders that are solvent soluble.  Hexamine is probe to sublimation at relatively low temeratures and used in fuel tablets, that might be kind of handy as it would be easy to remove even without sokvent.

Ulyimately I thi k I need to buy a bunch of different powders and get down to a lot of experimenting in the lab.  First, I need a lab. As of dec 7 I have signed a binding contract for a commercial bay that is a good start on a lab.  I have for some months is seed a storage unit so I have benches and may tools already.

However one of the major issues on the horizin is the sintering that occurs at high trmperature involved in casting steel and titanium etc.  Ultrasonic bonded solids can be expected to minimize this, as bondiing methods go, because thr small particles and small contact areas between particles that occur in colloid bonded powders are avoided.  The paricles are essentially already bonded with large contact areas, greatly reducing the ensuing shrinkage and distortion that will result from sintering.  Graphite is also one of the highest melting point materials we can easily work with, which implies relatively slow sintering rate.

Frequently in trying to design a promising apparatus, I encounter the problems the system is meant to solve.  Making custom parts is expensive and time consuming.

1-6 of 6