Metal deposition print head

From RepRap
Jump to: navigation, search

This page is the nucleus on this wiki for development of a metal depositing print head to go along with an XYZ Cartesian robot or robot arm.

The big picture

Machines are the main physical tools that allow material wealth creation. Good machines require precisely shaped metal parts, made to dimensional tolerances in the 10-30 micron range (roughly).

The MetalicaRap is moving forwards but there are many roads to glory :). Another particularly interesting and promising method is to form the metal object roughly and slightly oversized (near net shape) using a solid freeform fabrication method, and then machine the surfaces of the part that must be precisely shaped using a traditional subtractive method like EDM or milling, either after or during freeform fabrication.

While there do not seem to be any open precision cnc milling or EDM machines, the know how to make them is widely available, so a serviceably precise Open EDM or milling machine can certainly be made. Development of a suitable open freeform fabrication method suitable will enable the afor mentioned type of printer.

Such a machine would be relatively easy to have largely self replicating because it can print precision metal parts like a leadscrew, gears, linear journal bearings and fittings as well as snap together structural members, which is the bulk of the mechanical and structural components of such a machine. It could also print most of the parts for a Lister engine, hammer drill, a car (if the print volume was big enough), refrigerator etc.

Secondly, such a printer would probably have to already print in 2 different materials at least: a support material and a metal. Then, you could add an FDM head to the printer, allowing plastics to be printed onto the metal part as it is being made (within limits of not being exposed to excess heat). You would then have a multimaterial printer that could print the parts for even more of a car etc. Additional heads could be added like a head that deposited ceramic with the use of a diode laser, allowing you to produce parts mostly finished (including windings) for a stepper motor, most of an alternator, etc. Speed could be increased simply with multiple heads.

Existing freeform fabrication processes

See Comprehensive_search_for_full_strength_material_printers#Additive_processes_that_might_be_adaptable_to_printing See also the existing printer sections in that document. Ultrasonic consolidation is another particularly interesting one.

Proposed approaches

See http://en.wikipedia.org/wiki/List_of_welding_processes . Essentially we are adding a small bit of metal to a much larger bit, ideally consolidating it into a perfectly contiguous mass, which is very similar to welding with filler, so that is what most processes will draw on.

There is so called "micro welding" which involves units that may be a suitable size. They might be adapted as a convenient starting point for the gun or power supply etc.

Methods of adding material indiscriminately like metal vapor deposition or Chemical Vapor Deposition or plasma sputtering could also be used and then followed by subtractive machining.

==Wire and induction heating==


Wire as a feedstock is a handy way to go, easy to transport and, for the prototyping process easier to buy and handle. Under atmospheric pressure argon.

Primer for uninitiated: Remember current is induced in such a way that it produces a field that counteracts the changes in the primary magnetic field. In the diagram of magnetic field lines, with the higher density of magnetic field lines indicates the areas that are heated more. The higher the frequency, the less the heating tends to penetrate the surface of the inductively heated metal.

Induction heating could either use a plain, rod shaped air core solenoid or non conductive ferrite core magnet, with one pole of the magnet close to the metal surface, and modulated at relatively high frequency (10s of khz) to achieve an annular shaped heating area on the surface of a metal object (the melt pool).

Another way which could give a more uniform, circular heated area could be to have the magnetic field enter the surface over a small area and exit over a large area. Heating occurs at a useful intensity at the small pole and the magnetic circuit is closed a bit better which helps increase coupling into the pool. The coil could be water cooled.

The surrounding area could also be pre-heated with radiant heating using a nichrome or, since it is argon atmosphere, tungsten element, or the melt pool could be assisted a bit (how much? What sort of energy flux do we need here for various sized melt pools?).

Resistive heating

Arc

Try to basically scale down and make open one of the freeform fabrication ones methods that use wire under argon, or another arc welding method. Remember precision is not critical since there will be a machining step later.

Electron beam

Electron beam welding, basically. See documents at the main page on existing processes; nasa simply adapted a purchased electron beam welding system to use as a freeform fabricator.

The main problem is that their system used a vacuum of 5x 10-5 torr. While making a chamber for that level is doable, everything inside has to be compatible with that vacuum level too. Sensors, electronics, motors, lubricants, all become harder to integrate with the system inside the chamber when desirable, as buying off the shelf units specially made for high vacuum is probably very expensive. Could try to make our own but it's ultimately a real hurdle however you work it.


However strictly speaking only the electron gun itself needs to be under a high vacuum. There are many electron beam welding rigs that use low vacuum chambers and under normal atmospheric pressure (would still have to be argon though). Either a plasma window or even just a suitable arrangement of powerful pumps and narrow constrictions filled with gas is enough to keep the inside of the gun a vacuum while getting the electron beam out. A thin material though which the beam passes would probably not work due to overheating of it, or scattering or excess absorption of the beam? What if water cooled or it were moved (or the beam redirected) rapidly so the beam only passes through a given section very briefly? The thin material window can move relative to the electron gun easily; there are plastics and other materials which are sufficiently flexible and gas impermeable that they could be used to accommodate the motion while maintaining a seal. Or vacuum oil could be used with a metal to metal joint for a good seal.

Since we don't need a particularly well focused electron beam more research is needed on why NASA chose to use high vacuum in the entire chamber - money was no object maybe and they just didn't want to bother with plasma windows?

We could use the same gun as the MetalicaRap or a smaller one.

Ultimately the main benefit of electron beam for our purposes is probably only precision control of the energy into the melt pool? What else? The SMD and PTA FFF methods supposedly get full strength so it might not have much advantage but be more complex and expensive to prototype.

Summary of comparison between different approaches

I think we should steer clear of the high vacuum approaches anyway for this particular project, because of the cost and difficulty of prototyping and working with high vacuum. I think further research on the support materials will tell which is the best, but I am leaning towards induction if enough power can be coupled to the melt pool that way. GreenAtol 00:00, 14 August 2011 (UTC)

Argon supply

As mentioned in the reprap forum, Argon can be extracted from the air at a small scale economically and with relatively non expensive or complex machinery. So the cost or difficulty of obtaining Argon as a consumable is probably not a major issue ultimately, just the electricity cost. The documentation on the commercial FFF processes indicates high purity argon is desirable, (99.99%) and that 10 volumes of gas exchange in the chamber is typically needed to achieve good metallurgy. But even high purity Argon does not look too hard to make at a small scale. Seems unlikely that even if they start with 99.99% argon that the purity actually in the chamber after 10 exchanges will be near that, maybe though. Successive division operations can add (multiply) up pretty fast. Laminar flow rather than dilution flow in the chamber must be employed most likely.

Approaches include, may need to be combined or repeated for high purity:

  • Pressure swing adsorption.
  • Membrane separation (maybe with reverse osmosis membranes?)
  • Fractionation of air.

Existing open freeform fabrication processes or metal deposition heads

There are none known as far as I can tell right now.GreenAtol 00:11, 14 August 2011 (UTC)

Deposition speed

Obviously faster is better. However using thicker wire and more heat in the melt pool has limits and might have serious tradeoffs at some point. Using a row of identical deposition heads might be better. An array another step better still, as material could be deposited in a large number of places. An array of say 5 by 5 heads could be fixed relative to each other, and the motion required of the array to bring all the heads over the relevant points where material needs to be deposited could be either a scanning, raster approach, or there could be some performance enhancement possible using more complex motions that would be computed for each metal deposition layer. The melt pools could be sufficiently separated that they do not effect each other normally.

Fixturing

EDM is a nearly force free method. But milling entails significant force, though you can reduce the amount by moving the mill head more slowly.

So the object being fabricated needs to be firmly affixed to the rest of the printer somehow, the firmer the better. This is closely related to the support structure strategy. Options include, and these could be combined:

  • Have the bottom of the first print layer of metal stick to a ceramic baseplate, but not so much that it is impractical to remove. Could be removed by breaking the bond by differential thermal expansion or maybe just a good stiff whack. Limitations:
    • When the object first starts it's contact area with the base is small so it is poorly attached. Might have to be extra gentle with any milling, and/or maybe the process can be controlled to control the strength of the bond.
    • If the first print layer only touches the bottom of the chamber at a point or line you are probably out of luck, would have to use the support material approach along with in situ casting (see support section.)
  • The support material could be laid down in a single layer first as a sort of glue layer, if it had the right properties (has to stick to both metal and bottom of the print chamber of course and be adequately strong and temperature resistant WRT not burning and preferably not distorting (e.g. by melting) when the molten metal is deposited onto it). As the part is build up, the support material can be built up too, so it continues to provide a solid, rigid connection to the base. The corners of the build chamber could have vertical attachment points too, without interfering with the cartesian robot, which the support material can brace against. There may be a large amount of support material, if it can be reused in subsequent printings might as well go to town an fill any volume not filled with print metal with support material to provide quite solid support.
    • See support structure section.
  • There could be rods or pegs or wire sticking up from the base of the print chamber. Could be the same material as the print material. The metal could be deposited starting with welding it to the wire. Would provide grounding too for processes that require it.
    • Could either be 3 in a triangle shape and hold the object down just by being support pillars and through their own rigidity.
    • Could be only a single or 2 wires or pegs. The first bit of metal laid is welded to it/them, and then they are pulled downwards forcibly. This holds the object against the print chamber bottom and can continue to as the object grows if the support material surrounds it (or it supports it's self against the bottom of the chamber due to it's geometry of course). This could be combined well with the sticking to the bottom and corners of the chamber.
    • removing the pegs could be done by machining, or if the dimensioning is not critical there could be just a small mechanism built in to the printer base that cuts through the rod or wire, releasing the part. More wire or rod could then be fed up, and the printer can start a new cycle without manual intervention.


Support structures

For the uninitiated: When you print from the bottom up, layer by layer starting from the bottom, suppose you are printing a coat hanger, vertically, with the hook a the top. That is, viewing it in profile. The tip of the hook needs to be printed before it is connected in any way to the rest of the coat hanger. So it would need, in any case to be supported against gravity in the meantime. This bit may be called a "leader" before it is connected to the main object. Or you may be printing a large number of small unconnected objects... See fixturing and grounding sections too.

Patents reveal a lot of interesting information about possible materials and methods for support structures. Water soluble ceramics, structures that can be disintegrated with ultrasonic vibrations, high strength wax, and there is probably a lot more gold to be found.

If needed, more than one support structure could be used, each with their own strengths. So if a material was really good for attaching the metal to the base of the print chamber for instance for fixturing, but really slow to deposit.

Support structures perform the following functions/should have the following properties for the printers we are talking about here:

  • brace long thin vertical shafts and walls and parts that are not otherwise braced against the base of the print chamber (like the tip of the coat hanger hook before connected to the rest of the hanger, a leader) against the milling forces. This sort of entails being able to stick to the metal reasonably well and also be reasonably strong.
  • damp vibrations from the mill in long thin shafts and walls and floating points, again needs strength and stick
    • if it's a stout object with no overhangs and is not a leader it doesn't really need any bracing or support material since it can stand up by itself and would be affixed to the base of the print chamber. Those objects may still be useful so as a step of the dev process, printing these might make a good intermediate goal.
  • Act as a sort of in situ cast:
    • The subtractive machine tool will probably have a hard time in most cases machining the bottom surface of any perfectly horizontal overhangs of any significant length (a mill bit of the right shape or multi axis milling machine could help in many situations but there will still be a lot of geometries where it wouldn't be able to reach under there). In that case the support can be deposited, milled to the desired shape, and then the molten metal deposited on top. The precision would never be as good as milling the metal itself but it's better than nothing. Obviously this requires that it be heat resistant enough.
    • When printing a leader, the bottom of the leader would similarly be hard to machine, and could benefit from this.
    • Ultimately you still need to be able to machine under ledges at least a bit or you could only machine the top facing surface of any object which is an unacceptably severe limit on printable geometry. Suitable milling bits will probably suffice, or a "bit" with a small 90 degree gear built in, even, or the milling drive motor could be a water turbine or other quite small power plant (so it can reach in places reasonably), and the mill be a 5 axis mill.

A good support structure strategy, in this case, would also entail:

  • structures are easy to remove, preferably completely and not damage the surface. Metals are usually soluble in other metals. So if the support material was another metal like a low melting point alloy, it might damage the precisely machined surface anywhere the molten material touches, either when it is being deposited or removed, which is probably not acceptable. Maybe some alloys to choose from that do not have this problem with certain metals though.
    • In many cases it will have to disintegrate to be removed.
  • structures are manageable cost. Material may be recycled in many cases probably.
  • main problem is desirable material combined with practical deposition and removal.

Possible materials

Add examples of the water soluble ceramics, low melting alloys, high strength wax, also some sort of fast set gypsum (crystal structure absorbs moisture into it, dissolve to remove? Even if it takes some time to set that may be okay slows things down a bit but they are already pretty slow), materials like concrete which are sort of like epoxy (hardening occurs due to chemical reaction), the roughened stainless steel foil material pressed together one, salt or another material deposited in a solvent and the solvent evaporated, ultrasonic powder consolidation could work (powder plus pressure plus ultrasound), high speed droplets or bits of metal or deformable material fired at the surface (that droplet spray method sometimes used for making liners on engine cylinders?)(coudl be released with ultrasonic shaking maybe), maybe could be electrolytically deposited but probably pretty slow, vaporized and then deposits on there. Anything that liquiefies with cold rather than heat?

Thixotropic materials. Materials which sublimate at low atmospheric pressures? (could pressurize the build chamber too).

Fundamentally it is of course printing all over again, but the precision can be very low and it only has to work with one material with the right properties (although more might be useful of course especially if no perfect material which can also be easily printed with and removed can be found though here's hoping).

Grounding and supports

Unless the support material is a metal, or perhaps a metal trace from the main object is made out to any leaders may be problems grounding the leaders. The deposition method could maybe be designed so that the connection to ground is made right close to the melt pool, so if a GTAW like process were used there might be 2 arcs rather than 1, hitting the workpiece half millimeter apart and with current going down one and up the other. Or maybe better yet the wire could be the ground (the molten metal might be hard to maintain in continuity with the wire though). Induction could avoid this problem maybe.

Also obviously something needs to be figured out even for the main, non leader part, but just a wire sticking up from the bottom of the plate, to which the first bit of metal is welded should be fine.

Approaches that do not require grounding or have the current collector very close to the heated area:

  • Induction
  • Non-transfer plasma arc welding - basically super hot gas directed at the workpiece. Gas is heated by the arc, arc does not extend to the workpiece so it does not need grounding.
* Other methods using suitable gas flow patterns could be used to get heating only on small spots where desired e.g. have the gas exit from a small tube which is concentric with another tube which vacuums the gas and plenty of nearby atmospheric gas, back up, or even direct gas in a sort of cone shape towards the same area that the hot gas is being directed, keeping nearby areas exposed only to colder gas? The colder gas acting as a sort of wedge to get the hot gas away from the surface as soon as it has served it's purpose. Relative gas velocities probably have to be appropriate, but the surface of the molten pool would be concave so that might help direct the gas away from the surface again.  One issue might be splashing or vaporization of metal from the surface of the melt pool.  Might also be possible to heat the gas stream with a tungsten filament?
 
Conceivably the entire process could be be done at a fraction of atmospheric pressure, so that instead of arcs you can get stable gas discharges.  That may allow reduced electrode wear in the case of non-transfer plasma welding, or it could allow the two discharges to be more easily directed to the workpiece without difficult to manage instabilities? 
  • Chemically heated gas like oxyhydrogen welding - not as hot and probably not as precise as plasma, plus complications of hydrogen or oxygen or other stuff contaminating workpiece.
  • Heat the wire not the melt pool directly - need more info on welding, most likely problem is that not enough energy can be transferred (heat capacity of metal is too low)
  • Solid state welding like ultrasonic, EM pulse and explosive
  • electron beam welding uses only very low currents so it would be relatively easy to ground the object with a small metal probe that touches the metal, which probably doesn't work really with arc welding because the currents are so high. So high pressure ebeam then.


General information on welding

There is available on the web free copies of the book "Principles of welding: processes, physics, chemistry and metallurgy" which is pretty good. There is also "materials science and engineering callister" available, which has more information on welding and the materials science involved.

Finite element analysis/simulation of the weld process we are thinking of would be a good idea to see what sort of residual stresses and distortion is involved if we can get access to the software tools for it, although the commercial processes like LENS do not seem to have a problem with it and use melt pools of what are probably comparable size compared with what we are likely to.

The main difference though is that if the machining is to be done at each step distortion from the welding process after machining might be a real problem. Smaller melt pools might help, as might keeping the z axis separation between the machining plane and the metal deposition plane high, and precomputed compensation for any distortion that is expected. Ultimately ultrasonic consolidation may end up being the way to go as it is essentially distortion free.

feedback

a camera like http://www.xiris.com/products/sub-arc-viewing-camera.htm but smaller might be useful for prototyping, maybe made with a low cost camera plus filters. Feedback about melt pool temperature is not as critical when melting is done under atmospheric pressure as when under vacuum because the metal is not as prone to evaporation/boiling. A lot of information can be obtained from sensing the arc current, or by "ringing" the induction coil, etc. too so sensors per se may not be needed.

stress and strain

For the uninitiated: stress implies compression or tension in a material. Strain indicates actual flexing/distortion. So stress will cause strain because any force on an object will cause some flexing since nothing is perfectly rigid, and if you flex (strain) a part on purpose by say bending it with your hands, the strain implies that stress will appear in the object. So they always occur together and are closely linked.

For melting processes, the books on welding say substantial stress/strain is always substantial and due to both thermal expansion and change of volume when metal freezes. Stress may be relieved by heat treatment. Ultimately probably not a problem if you only wanted near net shape parts, but in out case we want finished parts so the amount by which the object is distorted *after machining* matters a lot. ,

Things may be reduced in severity by using smaller melt pool, and also by increasing the distance between the material addition zone and the final machining zone, keeping the entire object very hot during printing (reduces relative amount of expansion but then shape changes occur during cooling after printing) and maybe other means (use cold gas to get the heat out of the ibject as soon as possible after it leaves the melt pool?).

Solid welding processes don't have the problem nearly as badly since no melting/freezing and usually lower temperatures (especially ultrasonic).

Plasma or vapor deposition may be another low distortion material addition method


Stresses produced during melt deposition

Melt deposition has many advantages in terms of speed and the ability to deposit any alloy with full strength and other good material properties, but the thermal distortion is one of the main problems involved with freeform fabrication with this method. There are other metal deposition methods that might turn out to be plenty good and workable like sintering followed by CVD PVD, plasma deposition or electrochemical deposition, and the 2 cold spraying methods EM gun and gas nozzle type. Those require more research too and might turn out to be promising.


  • more on trying to get low or stress free parts (need info on degree of random changes that occur with thermal cycling or just heat/cool, and also the degree of predictability and randomness (or otherwise unpredictable changes) of changes in shape during the printing processes (the various processes) )
      • http://www.freepatentsonline.com/5866058.html method that anneals part as it is being made, heat it to above the cprrep point but below melting temp basically but plastic apparently , some basic info o ndistortion but may bot be the whole story as it does not factor in scan pattern etc. which we know makes a big diff , secondly does not adress the compensatability of this type of distortion which is obviously very high " Since most deposition materials change density with temperature, especially as they transition from a fluid to a solid, thermally solidifiable material rapid prototyping systems share the challenge of minimizing geometric distortions of the product prototypes that are produced by these density changes. Thermally solidifiable systems are subject to both "curl" and "plastic deformation" distortion mechanisms. Curl is manifest by a curvilinear geometric distortion which is induced into a prototype during a cooling period. The single largest contributor to such a geometric distortion (with respect to prototypes made by the current generation of rapid prototyping systems which utilize a thermally solidifiable material) is a change in density of the material as it transitions from a relatively hot flowable state to a relatively cold solid state.

For the simple case where an expansion coefficient is independent of temperature, the nature and magnitude of geometric distortion of sequentially applied planar layers can be estimated. Assume a linear thermal gradient dT/dz is present in a material when it is formed into a plate of thickness h in the z direction, and that the material has a constant thermal expansion coefficient α. The z direction is generally orthogonal to a support surface on which the plate is constructed. If the plate is subsequently allowed to come to some uniform temperature, it will distort, without applied stress, to form a cylindrical shell of radius r where: r=(α*dT/dz) -1 ( 1)

Curl C is defined as the inverse of the radius of curvature: C=1/r. An example of positive curl is shown in FIG. 1. Sequential layers of a thermoplastic material 104 are deposited on a base 102, using a moving extruder 106. As is typical in thermally solidified rapid prototypes, a series of layers are deposited sequentially in the z direction (i.e., the direction orthogonal to base 102), with the last layer deposited always having the highest temperature. Such an additive process typically results in a geometrically accurate part which contains a thermal gradient. As the part subsequently cools and becomes isothermal, the part distorts as a result of a curling of the ends of long features. "

  • search: no internal residual stress
    • from the EOS (DMLS system)site for the cobalt chrome material they sell to go with the printer, Stress relieving procedure:-

Stress relieving is done in a stress relieving furnace under argon atmosphere or in a vacuum furnace. The stress relieving sequence is as follows:- 1. ramp up to 650 °C in 60 minutes 2. hold for 3h 3. furnace heating power off and open the furnace door when temperature dropped down to approx. 400°C Annealing:- Specific properties can be modified by annealing the parts at various temperatures ranging from 650 to 850°C and for dwell times between 1 and 4h http://www.3trpd.co.uk/content/pdfs/ti64-titanium-alloy-eos-ti64.pdf for DMLS printed parts, does not specify precision well DMLS so would be 50-25 micron according to prior research

 uses a shaker of sorts low freq but can be ultrasonic

Librar Vibratory Stress Relief y REPRINT: VIBRATORY STRESS RELIEF: A FUNDAMENTAL STUDY OF ITS EFFECTIVENESS" Says library so migh tbe more where that came from

Patent No. PCT/GB88/00136 describe in detail the relevant apparatus for using the thermal tensioning technique to eliminate buckling distortions for butt welding of thin plates. "


  • relieving stress in metal and other material layers
    • extremely powerful ultrasound to releive stress? other ways to rleive stress or produce compressive stress to coiunteract the tensile stresses in the deposited metal layer other
    • other extreme shocks waves traveling through the devicee produced by exposives etc?
    • powerful em pulses?
    • deposite ions with an ion beam? almost any element can be used and so the alloy could remain good composition