Advanced Automation for Space Missions/Appendix 5F

There are two distinct classes of fabrication production machines in any general-product self-replicating system parts or "bulk" fabrication and electronics or microcircuit fabrication. Appendix 5F is concerned exclusively with LMF subsystems required for bulk manufacturing. Microelectronics production in space manufacturing facilities is considered in section 4.4.3 and is the subject of Zachary (1981); estimated mass of this component of the original LMF seed is 7000 kg, with a power draw of perhaps 20 kW to operate the necessary machinery (Meylink, personal communication, 1980).

5F.1 Overall Design Philosophy

The plausibility of both qualitative and quantitative materials closure has already been argued in appendix 5E. A similar line of reasoning is presented here in favor of a very simple parts fabrication system, to be automated and deployed in a self-replicating lunar manufacturing facility. To rigorously demonstrate parts closure it would be necessary to compile a comprehensive listing of every type and size of part, and the number required of each, comprising the LMF seed. This list would be a total inventory of every distinct part which would result if factory machines were all torn down to their most basic components - screws, nuts, washers, rods, springs, etc. To show 100% closure, it would then be necessary to demonstrate the ability of the proposed automated parts fabrication sector to produce every part listed, and in the quantities specified, within a replication time of T = 1 year, starting from raw elemental or alloy feedstocks provided from the chemical processing sectors.

Unfortunately, such a detailed breakdown and analysis probably would require tens of thousands of man-hours even for the simplest of systems. Not only is the seed not a simple system, but the present baseline design is not conveniently amenable to this sort of detailed analysis. Thus, a completely rigorous demonstration of parts closure is beyond the scope of the present study.

However, it is possible to advance a plausibility argument based upon a generalized parts list common to many complicated machines now in use in various terrestrial applications (Spotts, 1968; von Tiesenhausen, unpublished Summer Study document, 1980). Although machines designed for construction and use in space may employ radically different components than their terrestrial counterparts, to a first approximation it may be assumed that they will be comprised generally of the same kinds of parts found in commonplace machines on Earth such as bolt, nut, screw, rivet, pulley, wheel, clutch, shaft, crank, rod, beam, wire, plate, disk, bushing, cable, wedge, key, spring, gasket, seal, pipe, tube, and hose. If this is valid, then a showing that all parts classes in the general parts list can be manufactured by the proposed automated fabrication system may serve as a valuable plausibility argument in favor of parts closure for that system.

The achievement of a sound design which incorporates the advantages of maximum economy in manufacture and functional requirements of a part is dependent upon the designer's ability to apply certain basic rules (Yankee, 1979). There are four recognized rules, equally applicable to terrestrial factories and lunar replicating machine systems, as follows:


 * 1) Design all functional and physical characteristics for greatest simplicity. As a general principle, service life of a part is greatly increased when design of that part is both simple and sturdy ("robust"). Performance is more predictable and costs (money, build time, repair time) are lower for simpler parts.
 * 2) Design for the most economical production method. The particular production design selected should, if possible, be optimized for the part or set of parts the system must produce. The production of scrap (input/output ratio) is one valuable index by which optimality may be compared. This factor is relatively simple to evaluate where only one part is manufactured. In multipart production lines the problem is far more complicated, since each of the many parts may be expected to have dissimilar optima. Consequently, only the production of the entire system can be truly optimum.
 * 3) Design for a minimum number of machining operations. All types of costs are lower when fewer operations are required to produce a part according to specifications. The greatest savings result when the number of separate processing operations necessary to complete a part is reduced. Multiple operations which can be combined into fewer operations, or functionally similar parts requiring fewer production steps, should be changed in a design. "Needless fancy or nonfunctional configurations requiring extra operations and material" should be omitted from the design (Yankee, 1979).
 * 4) Specify finish and accuracy no greater than are actually needed. If a part will adequately serve its intended purpose at some lower level of accuracy of machining than is technologically possible, then cheaper, simpler production processes may be used which make closure easier to attain. The specification of needlessly close tolerances and an unreasonable degree of surface finish invariably results in a low part production rate, extra operations, high tooling costs, and high rejection rates and scrap losses (Yankee, 1979).

5F.2 Selection of Basic Production Processes

A wide variety of fabrication processes is available using current technology, each of which is optimum for the production of one or more classes of parts or in certain specialized applications (see table 4.17). From inspection of table 4.10 it is reasonable to conclude that there are perhaps only 300 fundamentally distinct fabrication techniques in widespread use today. Ultimately, the LMF factory in production phase may be called upon to perform many if not all of these functions. However, most may be unnecessary for initial system growth or replication. indeed, optimum seed design should permit maturation to adulthood in the minimum time with the fewest parts using the fewest machine operations possible.

The team concluded that four basic processes - plaster casting, vapor deposition, extrusion, and laser machining are probably sufficiently versatile to permit self-replication and growth. These four techniques can be used to fabricate most parts to very high accuracy. Plaster casting was selected because it is the simplest casting technique for producing convoluted parts as well as flat-surface parts, to an acceptable level of accuracy. (A number of alternatives have already been reviewed in app. 4B.) The laser machining tool can then cut, weld, smooth, and polish cast parts to finer finishes as required. Vapor deposition is the least complicated, most versatile method of producing metal film sheets to be used as the manufacturing substrate for microelectronics components, mirrors or solar cells, or to be sliced into narrow strips by the laser for use as wire. The extruder is used to produce thread fibers of insulating material, presumably spun basalt drawn from a lunar soil melt as described in section 4.2.2.

5F.3 Casting Robot

The casting robot is the heart of the proposed automated fabrication system. It is responsible for producing all shaped parts or molds from raw uncut elemental materials. The moldmaking materials it works with are of two kinds. First, the casting robot receives thermosetting refractory cement with which to prepare (a) molds to make iron alloy parts, (b) molds to make iron molds to cast basalt parts (but not aluminum parts, as molten aluminum tends to combine with ferrous metal), and (c) individual refractory parts. Second, the robot receives hydrosetting plaster of Paris with which to prepare (a) molds to cast aluminum parts and (b) substrates for the vacuum deposition of aluminum in sheets. According to Ansley (1968), small castings using nonferrous metals (aluminum, magnesium, or copper alloys) may be produced using plaster molds with a surface finish as fine as 2-3 μm and an accuracy of +/-0.1 mm over small dimensions and +/-0.02 mm/cm across larger surfaces (a drift of 2 mm over a 1 m$2$ area).

Traditionally, the plaster casting technique requires a split metal pattern in the shape of the object to be cast. This pattern is used to make a hollow mold into which molten metal is poured, eventually solidifying to make the desired part. Alternatively, patterns may be manually carved directly into the soft, setting plaster, after which metal again is poured to obtain the desired casting.

The casting robot should have maximum versatility. It will have access to a template library located within its reach, containing samples of each small or medium-sized part of which the LMF is comprised. If the SRS seed is designed with proper redundancy, it will use the fewest number of different kinds of parts and there will be large numbers of each kind of part. Assuming that on average there are 1000 pieces of each type of part in the original LMF architecture, then the total template library has a mass of only 100 tons/1000 = 100 kg and there are perhaps a thousand different kinds of parts (see below).

In addition, the casting robot is equipped with shaping and carving tools which can create any desired shape in the slowly hardening plaster. (Pure gypsum plaster hardens in 6-8 min after water is added, but this setting time may be extended up to 1-2 hr by adding lime, CaO, to the emulsion. Setting time is also temperature-dependent.) The shaping tools may represent perhaps 100 specific shapes and sizes and should also include at least a dozen "universal" carving instruments.

To make a given part, the robot searches its template library to see if it has a convenient pattern already in stock. If so, it uses the pattern to form the mold; if not, it uses its many tools to carve out a mold of the appropriate size and shape. Plaster of Paris is a hydraulic cement - it sets with the addition of water. Refractory cement is thermosetting and has to be heated to 1300-1400 K in a kiln to set the mold.

Water used to make the plaster molds cannot remain liquid in the lunar vacuum. Thus, the casting robot plaster system must be pressurized, probably with nitrogen gas to permit the pouring of molten aluminum. The triple point of water (the bottom end of its liquid phase) occurs at 608 Pa, but a 1.3&times;10$4$ Pa atmosphere (16 kg N2 to fill a 100 m$3$ working volume) prevents water from boiling off up to about 323 K.

Mass requirements for plaster molding are estimated by assuming that 10% of the volume of each mold contains a useful part (10% mold volume utilization). If the mean density of LMF parts (mostly aluminum) is taken as 3000 kg/m$3$, and the entire plaster mass is recycled once a day, then for a 100-ton seed the robot must have 2600 kg (0.91 m$3$) of plaster compound (gypsum, or calcium sulfate) on hand. To hydrate (set) this much plaster requires 483 kg of water, an amount of precious hydrogen already allowed for in LMF materials estimates presented in appendix 5E. Availability of sulfur is not a concern, since 2600 kg of plaster requires only 475 kg of S. Terrestrial plasters commonly have a small amount of strengthener added, but in the lunar application this substance should be designed to be recyclable or must be eliminated altogether.

Plaster casting is not the only way to make parts in a growing, self-replicating factory, but it is definitely one of the easiest both conceptually and in common industrial practice. Plaster methods are especially well suited for producing parts with hard-to-machine surfaces such as irregularly shaped exterior surfaces and in applications where a superior as-cast surface is important (Yankee, 1979). Plaster molded products commonly include aluminum match plates, cores and core boxes, miscellaneous parts for aircraft structures and engines, plumbing and automotive parts, household appliances, hand tools, toys, and ornaments. The technique is good for manufacturing parts requiring high dimensional accuracy with intricate details and thin walls (&gt;=0.5 mm). Castings of less than 0.45 kg and as massive as 11,350 kg have been made on Earth. Commercially, when compared to aluminum die casting, plaster mold casting is considered economical if 1000 parts or less are produced, although production runs up to 2000 parts may also be considered economical if the parts are especially complex.

Refractories. Refractories are materials which remain useful at very high temperatures, usually 1500-2300 K. They are employed primarily in kilns, blast furnaces, and related applications. In the lunar SRS refractories are needed as linings for drying kilns, roasting ovens, in the production of iron molds (to cast basalt parts) and iron parts, and also as material for special individual parts such as nozzles and tools which must operate at very high temperatures.

Refractories are usually, but not always, pure or mixtures of pure metal oxides. Tables in Campbell and Sherwood (1967) list the most important simple and complex refractory substances which LMF designers might choose. There are a few basic considerations, such as vapor pressure. For instance, although magnesia melts at 3070 K and has a useful operating temperature to about 2700 K in oxidizing atmospheres, it cannot be used in a vacuum at temperatures above about 1900 K because of volatization (Johnson, 1950). Similarly, zinc oxide volatizes above 2000 K and tin oxide sublimes excessively at 1780 K even in an atmosphere.

Refractory bodies are fabricated from pure oxides by powder pressing, ramming, extruding, or slip casting. The last of these is the simplest, but requires a very fine powder. This powder is normally prepared by ball milling. Steel mills and balls are used, and the iron is later separated by chemical means. For simplicity in LMF design, the iron alloy powder inevitably mixed with the milled product can be removed by magnetic separation.

High-alumina cements and refractories may be the best option for lunar manufacturing applications. Alumina is a maj.or product of the HF acid leach system in the chemical processing sector, and is capable of producing castable mortars and cements with high utility up to 2100 K (Kaiser, 1962; Robson, 1962). It will permit casting iron alloys, basalts, and low melting point metals such as Al and Mg. Unfortunately, it will not be possible to cast titanium alloys in this fashion, since in the liquid state Ti metal is very reactive and reduces all known refractories.

Alumina can be slip-cast from water suspensions. The oxide powder is first ball-milled as described above to 0.5-1.0 μm, then deflocculated by the addition of either acid (HCl) or base (NaOH), and finally the refractory body is developed by absorbing the liquid in a porous mold (plaster of Paris may be used with a base deflocculant). Gravity and hydrodynamic pressure of the flowing liquid produce a well compacted body of the suspended particles (Campbell and Sherwood, 1967). A fairly comprehensive review of alumina and alumina ceramics may be found in Gitzen (1966).

Metal alloys. A number of different metal alloys will be required for casting various parts and molds. Different alloys of iron may be chosen for the steel balls for ball milling, the basalt casting molds, and the individual part that might be comprised of steel or iron. Various aluminum alloys may be selected for parts, whereas pure metal is required for vapor deposition processes. Castable basalt may require fluxing but otherwise is a fairly straightforward melt.

Metallurgical duties are performed at the input terminus of the fabrication sector. Mobile chemical processing sector robot carriers dump measured quantities of metals and other substances into cold fabrication sector input hoppers (made of cast basalt and perhaps stored under a thin oxygen atmosphere to preclude vacuum welding). Mixing is accomplished by physical agitation, after which the contents are fed into a solar furnace to be melted. If net solar efficiencies are roughly the same as for the 5 kg capacity induction furnace (output 30 kg/hr) described in the MIT space manufacturing study (Miller and Smith, 1979), then about 30 kW of power are required which may be drawn most efficiently from a large collector dish roughly 6 m diam. There are at least three hopper/furnace subsystems required - a minimum of one each for iron, basalt, and aluminum alloys. Possibly another would be needed for magnesium alloys, and several more to forestall contamination between disparate batches, but three is the absolute minimum requirement.

Parts manufacturing. The construction Of a machine system as complex as a lunar SRS will require a great many individual parts which vary widely in mass, shape, function, and mode of assembly. If a complete parts list were available for the seed, then the manufacturing steps for each could be explicitly specified, precise throughput rates and materials requirements given, and closure demonstrated rigorously. Unfortunately, no such list is yet available so the team was forced to resort to the notion of the "typical part" to gain some insight into the performance which may be required of the casting robot.

Modern aircraft have about 10$5$ parts and weigh up to about 100 tons, for an average of 1 kg/part(Grant, 1978). The average automobile has 3000-4500 parts depending on its size and make, so the typical part weights perhaps 0.5 kg (Souza, personal communication, 1980). A study performed for General Motors concluded that 90% of all automotive parts weigh 2 kg or less (Spalding, personal communication, 1980). A design study by the British Interplanetary Society of a very advanced extrasolar space probe assumed a figure of 9 kg per typical part (Grant, 1978). Conservatively estimating that the typical LMF part is only 0.1 kg, then a 100-ton seed is comprised of roughly a million parts.

If most components may be made of aluminum or magnesium then the density of the typical part may be taken as about 3000 kg/m$3$, so the characteristic size of the typical part is (0.1/3000)$1/3$ = 3.2 cm. This result is consistent with Souza's (personal communication, 1980) suggestion that the average automobile part could be characterized as "roughly cylindrical in shape, an inch in length and half an inch in diameter." The casting robot must be able to cast all 10$6$ parts within a replication time T = 1 year. If the casting bay is only 1 m$2$ in horizontal extent, and only 10% of that area is available for useful molding, then each casting cycle can prepare molds for 0.1 m$2$ of parts. The characteristic area of the typical part is (0.1/3000)$2/3$ = 0.001 m$2$, and dividing this into the available area gives 100 parts/casting cycle as the typical production rate for the robot. To produce 10$6$ parts/year the casting robot must achieve a throughput rate or 10,000 cycles/year, or about 52 min/cycle. This in turn implies that the system must be able to carve or mold at an average rate of 30 sec/part. Since most parts should be simple in form or will have patterns available, this figure appears feasible. After the casting robot makes molds for the parts, the molds are filled with molten aluminum alloy. The metal hardens, the mold is broken, and the pieces are recycled back into plaster of Paris; the aluminum parts formed in the mold are conveyed to the laser machining and finishing station.

Very thin sheets of aluminum also are required in various applications, among them solar cell manufacture, production of microelectronic components, and solar furnace mirror surfaces. Extrusion, rolling, and direct casting were considered and rejected on grounds of lack of versatility and complexity. Vapor deposition, currently used in industry to apply coatings to surfaces and to prepare thin sheets of aluminum and other substances, was tentatively selected both because of its tremendous versatility (any curved surface may be coated) and because it is state-of-the-art technology. The major problems with the process in terrestrial applications are maintenance of the vacuum and high energy consumption, neither of which are factors on the lunar surface or in an orbital environment.

Plaster molds to be surfaced are passed to a laser honing station where they are finished to any desired accuracy, after which they move to the vapor deposition station and are coated with appropriate metals or nonmetals to the requisite thickness. The process is expected to proceed much as described by Miller and Smith (1979). The plaster mold is then removed and recycled, and the fabricated aluminum sheet is passed on to the electronic fabrication system or is sliced into wires by a fine cutting laser (Miller and Smith, 1979).

Mass throughput rates for this system appear adequate. Assuming that 10$4$ m$2$ of solar cells are needed for the original seed (Freitas, 1980) and that the casting bay is about 1 m$2$ in area, then for T = 1 year the required deposition rate to produce 0.3 mm thick aluminum sheet is r$d$ = (10$4$ m$2$ solar cells/year)(3&times;10$-4$ m thick/sheet)(1 sheet/m$2$)(1 year/5.23&times;10$5$ min)(10$6$ um/m) = 5.7 um/mm. State-of-the-art deposition rates attained for aluminum commercially are about 50 um/min (Miller and Smith, 1979), nearly an order of magnitude higher than required. (The above throughput rate would also be equivalent to 1 m/sec of 0.3 mm aluminum wire production if cutting and wrapping can keep pace with deposition). Cycling time is about 52 min/sheet. Following Johnson and Holbrow (1977), a heat of vaporization of 10$7$ J/kg for 10$4$ solar cells each made of 0.3 mm Al of density 3000 kg/m$3$ requires a continuous power draw of only 2.9 kW, which can be supplied by a small solar collector mirror 2 m in diameter.

A small number of LMF parts are expected to be made of cast basalt - fused as-found lunar soil perhaps with fluxing agent additives. Most parts will probably be aluminum because Al is an easily worked metal with high strength, low density (hence supporting structures need not be large), and relatively low melting point (hence is easily cast). The major advantages of basalt are its easy availability, its tolerance of machining, good compressive strength, and high density in some uses. Anticipated applications include machine support bases, furnace support walls, robot manipulator tools (to avoid vacuum welding), and other special parts where weight is not a problem. Because plaster fuses at 1720 K - very near the melting point of basalt - and loses its water of crystallization around 475 K, it cannot be used to make basalt castings. Iron molds cast from refractory templates are required; they may be reused or recycled as necessary.

Another principal application for basalt is as an insulating fiber. Spun basalt threads can be used to wrap electrical conductors to provide insulation, woven to produce "mineral fabrics" as filler to strengthen cements, shock-absorbing resilient packing material, filters and strainers for materials processing, or as thermal insulation or to prevent cold welding of metals (Green, unpublished Summer Study document, 1980). The technology for producing spun basalt products (Kopecky and Voldan, 1965; Subramanian and Kuang-Huah, 1979), basalt wool, and drawn basalt fibers (Subramanian et al., 1975) is well established commercially and customarily involves extrusion or simple mechanical pulling from a melt (see sec. 4.2 2).

Ho and Sobon (1979) have suggested a design for a fiberglass production plant for the lunar surface using a solar furnace and materials obtained from lunar soil (anorthite, silica, alumina, magnesia, and lime). The entire production facility has a mass of 111 metric tons and a power consumption of 1.88 MW, and produces 9100 metric tons of spun fiberglass per year. Assuming linear scaling, the production for the replicating LMF of even as much as 10 tons of fiberglass thread would require a production plant of mass 122 kg and a power consumption of 2.1 kW (a 2-m solar collector dish).

A small number of LMF parts will also be made of iron (from refractory molds) and refractory cements (carved directly from ceramic clay by the casting robot) in order to take advantage of the special properties of these substances. The total mass of such items is expected to be relatively low. Used refractory molds may be fed to the ball mill and recycled if necessary.

5F.4 Laser Machining and Finishing

The plaster casting parts manufacturing technique was chosen in part because of its ability to produce ready to use "as-cast" components. Thus, it is expected that the majority of parts will require little reworking, machining, or finishing. A small fraction, perhaps 10%, of all lunar SRS parts may require more extensive machining. A laser machining system was selected for this function in the LMF. The characteristic circumference of the typical part is 3.14(0.1/3000)$1/3$ or about 10 cm. If surface articulations cause an increase by a factor of ten in the total average path length that must be machined, then the mean operating speed of the laser system must be (10$6$ parts/year)(10% machinables)(0.1 m/part)(10 m path/m circum.)(1 year/8722 hr) = 11.5 m/hr. Table 5.16 compares the performances of several different types of lasers, and table 5.17 gives specific performance parameters for high-power gas lasers used in industry for welding (butt, lap, comer, and edge) and for cutting. Inspection of these values suggests that a 5-10-kW continuous-wave (CW) carbon dioxide laser should be able to weld and cut "typical parts" with characteristic dimensions up to 3 cm at the required throughput rate.

$a$Maximum thickness given here is for Type 304 stainless steel.

Laser cutting speeds typically are as much as 30 times faster than friction sawing (Yankee, 1979). Cutting accuracy is about 0.01 mm/cm under closely controlled conditions. All metals - including high-strength, exotic, and refractory alloys such as Inconel and titanium, as well as aluminum, stainless steel, and brass - and nonmetals such as diamond, ceramics, and plastics may be vaporized by laser beams. Hence, parts of these materials may be easily machined. Burr-free laser holes may be drilled as small as 10-100 μm. Lasers can also be used for pattern cutting, gyro balancing, insulation stripping, surface hardening, trimming, photoetching, measurement of range and size to 1 μm accuracy or better, scribing 5-10 μm lines on microelectronic wafers, flaw detection, marking or engraving parts, and impurity removal (e.g., carbon streaks in diamond). Laser beam machining is "especially adaptable and principally used for relatively small materials processing applications such as cutting, trimming, scribing, piercing, drilling, or other delicate material removal operations similar to milling or shaping" (Yankee, 1979).

Dunning (unpublished Summer Study document, 1980) has suggested a variety of space and lunar applications for laser machining, including flash trimming of cast basalt parts; engraving bar codes on parts to enable quick and accurate recognition by robot vision systems; drilling holes in workpieces an inch thick or less; internal welding of cast basalt joints, pipe, and structural members; impurity removal from lunar-produced semiconductor chips; cutting operations on gossamer structures (Brereton, 1979) in orbit; and case hardening of cast basalt or metal parts. Dunning has also suggested two potential major problems associated with the use of lasers in the context of a selfreplicating, growing lunar manufacturing facility: (1) the need for gas jets, and (2) the requirements of closure.

In normal industrial usage, vaporized workpiece material is carried away by a gas jet, usually oxygen (Yankee, 1979). The gas serves three functions: (1) to oxidize the hot working surface, decreasing reflectivity, (2) to form a molten oxide (i.e., the metal "burns") which releases a large fraction of the useful cutting energy, and (3) to remove slag and hot plasma from the path of the beam. There is no problem maintaining a moderate-pressure O$2$ atmosphere around the laser work area, as the beam penetrates air easily. In this case the usual gas jet can still be used. Or, the laser could be placed outside the pressurized working area, shooting its beam through a transparent window. If pressurization must be avoided, laser machining can be done entirely in vacuum and the ionized plasma wastes removed by a magnetic coil following the cut or weld like an ion "vacuum cleaner." However, it is estimated that up to 80% of the laser cutting energy comes from the exothermic oxidation reaction, so in this latter case laser energies would have to be on the order of five times the value for the equivalent O$2$-atmosphere machining.

The problem of closure is even more critical in a replicating autonomous remote factory. The materials closure problem is solved in large measure by resorting to CO$a$ gas laser technology. This gas is available in limited quantities on the Moon, whereas materials for solid state lasers such as yttrium, ruby, garnet or neodymium are generally very rare (although Dunning has suggested that spinel, which is plentiful on the Moon, might be substituted for garnet). Quantitative materials closure may be argued as follows. A typical CO$2$ laser uses three gases for high-power operation - carbon dioxide to lase, nitrogen to sustain the reaction, and helium for cooling because of its excellent heat conducting properties. Since oxygen is plentiful, the three limiting elements are C, N, and He. From appendix 5E, the LMF in one year can produce 400 kg C, 400 kg N$a$, and about 40 kg inert gases (at least 90% of which is He). This is sufficient to make 747 m$a$ (33,300 moles) of CO$a$, 320 m$2$ (14,300 moles) of N$2$ and 224 m$2$ (10,000 moles) of He, at STP. Even if the laser machining device requires several hundred moles of these gases (a few thousand liters at STP), still only a few percent of available LMF stocks of these elements need be diverted for this purpose, a negligible resource drain.

The problems of parts and assembly closure cannot be answered satisfactorily at the present time. However, it is often asserted that machining the laser end mirrors to high accuracy may be a major roadblock to automated manufacture of lasing devices. Nazemetz (personal communication, 1980) has pointed out that a laser is accurate enough to surface a rough-hewn mirror to the accuracy required for its own construction. In a pinch, concave mirrors could be hewn from solid metal or basalt blanks simply by sweeping the laser beam radially across the disks, applying higher power nearer the center so more material volatizes there, thus creating a perfect spherical or parabolic surface gradient. There appear to be no major unresolvable difficulties associated with the use of lasers in an autonomous lunar manufacturing facility.

After parts leave the laser machining station they may require some slight further treatment such as annealing or coating to prevent cold weld, though this latter function may be unnecessary if laser welding takes place in an oxygen atmosphere (a thin layer of metal oxide prevents the vacuum-welding effect). Once fabrication is completed each part may have one of three possible destinations: (1) assembly sector, where the part is given to a mobile robot for transport to wherever it is needed, (2) parts warehouse (which serves as a buffer supply of extra parts in the event of supply slowdowns or interruptions), where the part is taken to storage by a mobile robot, or (3) fabrication sector, when more fabrication must be performed upon an already manufactured "part" (e.g., solar cell aluminum sheets), where a mobile robot carries the part to wherever it is needed in the fabrication sector. A general flowchart of the entire automated parts fabrication process appears in figure 5.17.

5F.5 Parts Fabrication: State-of-the-Art

In the operation of any general-purpose fabrication machine (mill, lathe, laser machining system, casting robot, there are seven distinct functions which must be performed either manually or automatically, according to Cook (1975):


 * 1) Move the proper workpiece to the machine,
 * 2) Load the workpiece onto the machine and affix it rigidly and accurately,
 * 3) Select the proper tool and insert it into the machine,
 * 4) Establish and set machine operating speeds and other conditions of operation,
 * 5) Control machine motion, enabling the tool to execute the desired function,
 * 6) Sequence different tools, conditions, and motions until all operations possible on that machine are complete, and
 * 7) Unload the part from the machine.

Traditionally all seven operations were performed by the human operator. The development of numerical-control (N/C) machining relieved human operators of the need to manually perform step (5), and automatic tool-changing systems supplanted step (3). Although most modern computer-controlled machining systems have "a finite number of tool-storage locations - 24, 48, or 60 tools, for example - the number that could be built into a system runs into the thousands" (Gettleman, 1979). If the seed is comprised of about 1000 different kinds of parts, each requiring a template pattern for the casting robot, Gettleman's estimate for N/C machine tooling makes plausible the satisfaction of this requirement by extensions of current technology. Adaptive control of N/C machine tools, with sensors that measure workpiece and tool dimensions, tool application forces, vibration and sound, temperatures, and feed rates to optimize production have already been developed (Nitzan and Rosen, 1976) but will require further improvements to achieve the kind of generalized capability required for a lunar SRS.

The next logical developmental step is the design of a completely computer-managed integrated parts manufacturing system. Cook (1975) describes such a system developed and built by Sunstrand Corporation. One version in operation at the Ingersoll-Rand Company is used primarily for fabricating hoists and winches, while another at the Caterpillar Tractor Company is used for making heavy transmission casing parts (Barash, 1976). As of 1975 there were about ten similar systems in operation in the U.S., Japan, Germany, and the U.S.S.R. (Barash, 1975).

The Ingersoll-Rand system consists of six NIC tools - two 5-axis milling machines, two 4-axis milling machines, and two 4-axis drills - arranged around a looped transfer system as shown in figure 5.42. Machining operations include milling, turning, boring, tapping, and drilling, all under the control of an IBM 360/30 central computer. At any given time about 200 tools are in automatic toolchanging carousels, available for selection by the computer, although about 500 are generally available in the system. The computer can simultaneously direct the fabrication of as many as 16 different kinds of parts of totally different design which are either being machined, waiting in queue to be machined, or are in the transfer loop. The entire system is capable of manufacturing about 500 completely different parts. During each 12-hr shift the system is run by three human operators and one supervisor. It is calculated that to achieve the same output using manual labor would require about 30 machines and 30 operators. Finally, the circular pallets used to present parts to each control station have maximum dimensions which fit inside a 1-m cube, exactly the scale discussed earlier in connection with the casting robot.



Another major advance is the variable-mission manufacturing system developed by Cincinnati Milacron Inc. This system not only has the general character of computer managed parts manufacture seen in other systems but also provides for the processing of low-volume parts at higher rates than those which can be achieved with more conventional N/C machines. For instance, an ingenious five-axis "manufacturing center" automatically changes clusters of tools mounted on a single head so that a number of operations can be performed simultaneously by means of a novel scheme of handling workpieces from above, the Cincinnati Milacron system provides efficient management of coolants and chips, together with easy access for inspection and servicing (Cook, 1975).

The Japanese have been most aggressive in pursuing the "total automation" concept. During 1973 through 1976 their Ministry of International Trade and Industry (MITI) supported a survey and design study entitled "Methodology for Unmanned Manufacturing" (MUM) which forecast some rather ambitious goals. The MUM factory was to be operated by a 10-man crew, 24 hr/day, and replace a conventional factory of about 750 workers. The factory will be capable of turning out about 2000 different parts at the rate of 30 different parts (in batches of about 1-25) per day, which will be inspected and assembled to produce about 50 different complex machine components such as spindle and turret heads, gear boxes, etc. Machining cells, based on the principle of group technology, will be controlled by a hierarchy of minicomputers and microcomputers, and will receive workpieces via an automated transfer system. Each machine cell will be equipped with inspection and diagnostic systems to monitor such useful parameters as tool wear, product quality, and the conditions of machine operation. Assembly cells, much like the machining cells, will be equipped with multiple manipulators fashioned after present industrial robots, together with an automated transfer system for movement of assemblies (Nitzan and Rosen, 1976). One ultimate program goal, explicitly stated, was to design a system "capable of self-diagnosis and self-reproduction ... [and] capable of expansion" (Honda, 1974).

Following this initial study, MITI in 1977 initiated a 7-year national R&amp;D program at a funding level of 12 billion yen (about $57 million) to develop, establish, and promote technologies necessary for the design and operation of a "flexible manufacturing system complex," a prototype "unmanned" factory to be built sometime in the mid-1980s (Ohmi et al., 1978). The technologies currently receiving emphasis include:


 * Optimum design and integrated control of manufacturing systems including blank fabrication, machining and assembly,
 * Flexible machining for mechanical parts and components,
 * Enlargement of the flexibility of blank fabrication,
 * Enlargement of the applicable area of automatic assembly and automatic transfer,
 * Application of high-power (20 kW) CO$2$ lasers to metalworking,
 * Automatic diagnosis of manufacturing facilities to detect malfunctions, and
 * Planning and production management to optimize system operation.

MUM presently is being pursued vigorously by three government research institutes and 20 private companies, and is being managed by the Agency of Industrial Science and Technology of MITI (Honda et al., 1979).

The original forecast was that MUM technology would go into operation sometime during the 1980s. At a conference in Tokyo in September of last year, Fujitsu FANUC Ltd., a leading international manufacturer of numerical control (NIC) machining equipment, announced its plans to open a historic robot-making factory near Lake Yamanaka in Yamanashi Prefecture in late November. At the plant, then still under construction, industrial robots controlled by minicomputers would produce other industrial robots without major human intervention save minor machine operation and administrative tasks. The plant is the first "unmanned" factory in the world machinery industry. producing robots and other equipment worth about $70 million in the first year of operation with only 100 supervisory personnel. In 5 years the plant is expected to expand, perhaps with some of the robots it itself manufactures, to a $300 million annual output with a workforce of only 200 people, less than a tenth the number required in ordinary machine factories of equivalent output. The mainstay products are to be various kinds of industrial robots and electronic machines. A spokesman said that FANUC's fully automated system is suitable not only for mass production of a single product line but also for limited production of divergent products (IAF Conference, 1980).

An automated plant in which robots make robots is a giant first step toward the goal of a practical self-reproducing machine system. When a factory such as the FANUC plant can make all of the machines and components of which it itself is comprised, its output can be specified to be itself and thus it can self-replicate. It appears likely that the automation technology required for LMF fabrication and assembly operations could become available within the next 10-20 years, given adequate funding and manpower support targeted specifically to the development of such a system.

5F.6 Automation of Specific LMF Systems

It is useful at this point to consider the automation potential of specific LMF systems. Most critical are the casting robot and the laser machining system, but several other subsystems will also require automation.

Casting Robot Automation

There are two potential precursor technologies to the general-purpose casting robot described in section 5F.3, in addition to established robotics devices such as the Unimate 4000 that produces lost wax ceramic molds for use in investment casting (Moegling, 1980). One of these lines of development has been in the field of precision machining, the other in the area of art and sculpturing.

Engraving and tracer milling are well established machining techniques. These machines use high-speed spindles mounted on pantograph mechanisms guided by master patterns which permit the cutting tools to be guided from an original which may be larger or smaller than the workpiece. The original pattern may be wood, plastic, or metal; the operator follows it with a guide and the machine faithfully reproduces each motion - but enlarges or reduces it as desired (Ansley, 1968).

Modern machines work in three dimensions and can be used for very intricate carving in metal from arbitrary solid originals. A contour milling machine developed by Gorton Machine Corporation uses numerical control to replace entirely the master pattern and the human operator (Ansley, 1968). A skilled technician can preprogram the complete machining cycle for any given part. The Lockheed CAD/CAM system (see below) permits still more sophisticated computerized design and parts fabrication. It seen but a few conceptually simple steps from this level of technology to that required for a "universal" contour-carving device like the casting robot. Such a system will require vision system, excellent tactile sensing, an automatic tool-changing and pattern-changing capability, and development of an automatic feedstock handling system for metal, gases, and refractories.

Another possible precursor technology to the casting robot may be found in the area of artistic sculpting, otherwise known as "three-dimensional portraiture" An excellent summary of 19th-century attempts to construct machines able to automatically size and shape a human head for personalized sculptures has been written by Boga (1979). In the last 10 years two very different descendants of the 19th-century efforts to produce sculpted likenesses (thus bypassing the creative artist) have been spawned. The first of these is modern holography techniques, which permit the generation of 3-D images using laser beams and, more recently, white light sources.

The second technology, often called "solid photography," requires that the human model pose in front of eight cameras shooting simultaneously from different angles. Linear patterns of light are projected onto the subject's face and all three-dimensional information is coded by the cameras. The coded films are then read by an optical scanner which converts the code into digital information which is processed by a computer to produce an accurate surface map of the person or object. This map is then translated into a series of cutting instructions which are passed to two cutting instruments.

In the system operated by Dynell Electronics Corporation of Melville, New York, instructions are first passed to a "coarse replicator" which rough-hews the shape of the human head in paralene wax (high melting point) in 90° sections. After about 30 min, the rudimentary carving is completed and is passed to the "fine-cut replicator" which is also computer-controlled. This time, instead of a single rotating bit, the tooling consists of 20 rotating blades that finish the work to a very high accuracy in about 40 min of work. Human hands are used only for touch-up of very fine details or for imparting skin-like smoothnesses; witnesses to the procedure are impressed with the results - excellent representations of eyebrows, locks of hair, creases, even moles (Field, 1977). Clearly, the Dynell automated sculpting system is not too distant from the casting robot, conceptually or technologically. If treated as a serious item for further development, it is likely that casting robot technology could be ready in a decade or less starting from the current state-of-the-art.

Laser Machining System Automation

Nonlaser spot welding has been a standard automated industrial technique for many years. Welding robots at Chrysler's Hamtramck assembly plant put uniform spot welds on parts assemblies with positional accuracy exceeding 1.3 mm. Typical operation includes a sequence of 24 welds on four automobile assemblies at once (Tanner, 1979). One of the largest and most fully automated welding lines in the world operates at Volvo's Torslanda plant in Gothenburg, Sweden. The new welding line consists of 27 Unimate robots which replace 67 workers with 7. The installation is fully automated, including loading and unloading stations, intermediate assembly of all automobile body parts, lining, and clamping preparatory to welding. The line does a total of 754 spot welds per assembly, and each Unimate is directed by 2-8K programmable controller computers (Mullins, 1977). Kawasaki Unimate robots have been applied to are welding of motorcycle flames and automobile rear axle housings (Seko and Toda, 1974). Accuracy in are welding is more difficult to achieve than in spot welding, but apparently much progress has been made in this area.

Nonlaser machining is also highly automated. The generalized machining center can perform a number of functions in typical operation including milling, drilling, boring, facing, spotting, counterboring, threading, and tapping, all in a single workpiece setup and on many different surfaces of the workpiece (Gettleman, 1979). A numerical-control machine operated by the Giddings and Lewis Machine Tool Company has an automatic tool changer with 40 tools. It machines all sides of a workpiece with one setup. (Setup time is usually 50-90% of total machining time, and a typical part might normally require a dozen setups or more, so this is a substantial savings.) A machined block requiring 174 separate operations can be completed automatically in 43 min; the former method required 4 machines with 3 operators and took 96 min to finish the part. Piggott (personal communication, 1980) estimates that a "typical part" weighing 0.1 kg will require about 20 machining operations. If 10% of all LMF parts must be closely machined after casting, a single Giddings N/C robot could perform all 2,000,000 necessary machining operations in just 0.94 year. Since several such robots could be available in the early LMF, this item is noncritical.

A more sophisticated methodology (Luke, 1972) is used in the Lockheed CAD/CAM system. In this system, the user designs a part of arbitrary shape in three dimensions on an interactive computer-driven TV console. This description is processed to yield a series of machine operations and is then passed to a set of 40 sophisticated N/C machines which make the part "from scratch" out of feedstock supplied at one end. On the average, parts are machined correctly five out of every six tries.

If all LMF parts had already been designed and placed in memory, a shop in space using the Lockheed system could manufacture each of the 1000 different SRS parts. With the addition of pattern recognition software capable of recognizing any part presented to a camera eye, in any physical condition (e.g., rotated, broken, partly melted, partly obscured) (Perkins, 1977), and a simple goal-setting command hierarchy, the Lockheed system might be able to recognize and repair damaged parts presented to it randomly.

The purpose of describing the above nonlaser welding and machining systems is to suggest that laser machining should be equally automatable because the laser may be viewed as another modality for delivering heat or cutting action to a workpiece. Any nonlaser automated welding/machining technology in principle may be modified to accept a laser as its active machining element.

Lasers already have found many automated applications in industry. Computer-driven lasers presently perform automated wire-to-terminal welding on relay plates for electronic switching circuits (Bolin, 1976). There are automated laser welding lines for manufacturing metal-enclosed gas-protected contacts for telephone switchgear (Schwartz, 1979). A computer-controlled laser welding system at Ford Motor Company allows welding parameters for a number of different automobile underbody designs to be stored in the central memory and retrieved as required for seam welding body-pans (Chang, personal communication, 1978). In the garment industry, the cutting of patterns from single-ply or multilayer stacks of fabrics is easily fully automated and rates of up to 61 m/min have been achieved (Luke, 1972; Yankee, 1979). Flash trimming of carbon resistors has been successfully automated. Automated marking and engraving (with alphanumeric characters) is another application of computer-guided lasers (Yankee, 1979). Numerous other laser applications have already been put into operation (see sec. 5F.4) but are not yet automated. Lasers for many automobile body assembly tasks are impractical today because the component metal pieces to be welded,which are stamped metal sheet, are too inaccurate to permit a close enough fit for laser welding to be feasible - though, according to Schwartz (1979), "this situation may change gradually in the future."

Lunar seed lasers should be able to operate at many different power settings, preferably spanning a broad continuum. Precision machining of liquid- and air-tight valves, laser mirror surfaces, and various other small intricate parts will demand the closest scrutiny of the rate at which energy is delivered to the workpiece. Lasers may also be used for super-accurate ranging and sizing measurements, which require an ultralow power capability as well as sophisticated optics, timing, and data processing systems. Automation of the LMF Laser Machining System will require close computer/mechanical control to perform each of the seven basic machining steps described earlier in section 5F.5.

Some consideration should also be given to the architecture of beam delivery to the workpiece. Laser power may be transmitted directly, in which case the entire laser assembly must be swiveled as various operations are performed. One alternative is to use a system of lightweight movable mirrors to angle laser energy in the desired direction to impact the workpiece. Reflectivities up to 0.86 for aluminum on glass would give an absorbed power density of 14 to 140 W/cm$2$ for a 1-10% efficient 10 kW laser beam with a 1 cm$2$ cross section. This heating may be reduced by at least an order of magnitude by "jiggling" the mirrors along their plane to spread the beam impact spot over a wider area while maintaining precise directional control. Another possible solution is to locate a high power laser in some central location and convey the beam to its destination via large fiber-optic light pipes. There are possible materials closure problems with fiber-optics, and absorbed energy may damage or destroy the glass, but this alternative offers many interesting opportunities and cannot be logically ruled out.

The team recognizes that lasers may not be the optimum technology for an autonomous replicating lunar facility. Their inclusion in the present design is intended as a heuristic device to illustrate, not unequivocally select, a particular option. For example, industrial experts in manufacturing technologies are split over whether lasers or electron beams are generally superior or more versatile, e.g., Schwartz (1979) favors lasers and Yankee (1979) favors e-beams. The MIT study group selected electron-beam cutting over lasers because "lasers are less efficient and require more maintenance and repair than EB guns" (Miller and Smith, 1979), a conclusion not adequately documented in their final report.

Nor is it absolutely clear that conventional machine tools such as mills, lathes, or drills are unsuitable for use in space. The problem most often cited in this context is that the tool bit and workpiece may vacuum weld during machining. However, cold welding is known to occur only between identical metals or between those with very similar crystallographic characteristics (such as aluminum and magnesium). Steel, for instance, will not vacuum weld to aluminum. Neither will any metal part cold weld to cast basalt.

Further, ceramic cutting tools have recently been developed which have increased the cutting speeds of mills and lathes dramatically. When tungsten carbides were introduced in 1929, cutting speeds quadrupled to 100 to 200 m/min. Since the 1950s, ceramic and other cemented oxide (alumina) and refractory tool materials such as nitrides and borides have been successfully employed in achieving cutting rates of 300 m/min and higher (Ansley, 1968). Ceramic tools will not cold weld to anything.

A more critical problem would seem to be the seizing of internal machine components, rather than vacuum welding between tool and workpiece. This difficulty could perhaps be surmounted by bathing enclosed machinery in lubricants, a light oxygen atmosphere trapped by airtight seals, or by using basalts or ceramics to construct or merely protectively coat internal machine moving parts.

Automation of Other Systems

The remaining subsystems within the parts fabrication sector must also be automated for full LMF autonomous operation. These subsystems include:


 * 1) Kilns and metallurgical furnaces: The extraterrestrial fiberglass production system using solar energy, designed by Ho and Sobon (1979), is designed to be automated. This system includes melting and drawing operations. According to the authors, "the systems will be automated, but minimum manpower will be required for maintenance. For the lunar plant, maintenance will be required at the beginning of each lunar day to begin the drawing process."
 * 2) Basalt threads: The system of Ho and Sobon will be automated. Also, a series of eleven specific steps which a manufacturing robot such as a Unimate must perform in order to completely automate the thread-drawing procedure is given in appendix 4D.
 * 3) Wire wrapping: An automatic insulation wire-wrapping machine has been described in some detail by Miller and Smith (1979).
 * 4) Sheet metal and cutting operations: Miller and Smith (1979) discuss in some detail aluminum ribbon and sheet operations. Vacuum vapor deposition as a fabrication technique is also described in Johnson and Holbrow (1977). These will be at least partially automated.
 * 5) Refractory and cement production: Ansley (1968) has described a concrete batching plant equipped with electronic controls permitting the selection of some 1500 different formulas and which give twice the output of manually operated plants. Batches are prepared by inserting a punched card into a reader to specify the formula to be used, and the system does the rest automatically if adequate materials have been supplied.
 * 6) Ball mills and magnetic purification: These are standard automated technologies, assumed available in space processing models provided by O'Neill (1976), Phinney et al. (1977), and others.

5F.7 Sector Mass and Power Estimates

In lieu of a complicated breakdown of fabricator sector component subsystems with detailed analysis of each, table 5.18 illustrates a more practical approach. This information was assembled from various sources and gives typical masses and power requirements for parts fabrication facilities in previous studies.

The nominal annual output of the original lunar seed is 100 tons/year. Using the most extreme machine productivity values given in table 5.18, fabrication sector mass may range from 137 kg up to 20,400 kg. A similar comparison with the power requirements values gives a range of 0.3-345 kW for sector energy consumption. The upper ranges of these estimates are probably most appropriate in the replicating lunar factory application.

5F.8 Information and Control Estimates

Even in the absence of a detailed analysis of the necessary control operations, it is obvious that the complete description of all parts will dominate computer memory requirements. Since each typical part has a characteristic surface area of 10$3$ m$2$, then if the surface of each is mapped to 1 mm$3$ resolution per pixel, each part will require 1000 pixels for complete coverage. Each pixel must identify three position coordinates, materials used, machining operations to be performed, etc. If 100 bits/pixel is adequate, then roughly 10$2$ bits/part are required in memory for a total of 10$3$ bits of storage for all 1,000,000 parts in the original lunar seed. This crude estimate is intended as a combined total for description and operation of the system.

Subsystem control hardware is likely to use vastly less computer capacity than this. The entire Sundstrand integrated parts manufacturing line is managed by an IBM 360/30 central computer with microcomputers driving each robot station. While some tricks might be employed to reduce redundancy (such as "chunking" large similar areas), more convoluted surfaces will require extra description. It is likely that the main driver will be the requirements for parts description.

5F.9 References

Acharekar, M. A.: Lasers Finding Growing Use in Industrial Welding. Welding Engineering, December 1974, pp. 9-11.

Ansley, Arthur C.: Manufacturing Methods and Processes. Chilton Book Company, Philadelphia, 1968. Revised and enlarged edition.

Barash, M. M.: Optional Planning of Computerized Manufacturing Systems. Proc. 3rd NSF/RANN Grantee's Conference on Production Research and Industrial Automation, Cleveland, October 1975.

Barash, M. M.: Integrated Machining Systems with Workpiece Handling. In Proc. 6th Intl. Symp. Industrial Robots and Industrial Robot Technology, University of Nottingham, March 1976, International Fluidics Services, Kempston, 1976, pp. H4/29-36.

Bogart, Michele: In Art the Ends Don't Always Justify Means. Smithsonian, vol. 10, June 1979, pp. 104-108, 110-111.

Bolin, S. R.: Bright Spot for Pulsed Lasers. Welding Design and Fabrication, August 1976, pp. 74-77.

Brereton, Roy G., ed.: Discussion Meeting on Gossamer Spacecraft (Ultralightweight Spacecraft): Final Report, JPL Publication 80-26, Jet Propulsion Laboratory, Pasadena, California, 15 May 1980. 174 pp. NASA CR-163275.

Campbell, Ivor E.; and Sherwood, Edwin M., eds.: High-temperature Materials and Technology. Wiley, New York, 1967.

Cook, Nathan H.: Computer-Managed Parts Manufacture Scientific American, vol. 232, February 1975, pp. 22-29.

Field, Roger: Computerized Cameras, Knives Sculpt Quickly. Science Digest, vol. 81, April 1977, pp. 77-78.

Freitas, Robert A., Jr.: Baseline Reference Design for a Growing, Self-Replicable, Lunar Manufacturing Facility. Summer Study Document, 8 July 1980. 56 pp

Getdeman, K. M.: Fundamentals of NC/CAM. Modern Machine Shop 1979 NC/CAM Guidebook, 1979.

Gitzen, W. H: Alumina Ceramics. U.S. Air Force Rept, AFML-TR-66-1621, 1966.

Grant, T. J.: The Need for On-Board Repair. In Project Daedalus - The Final Report on the BIS Starship Study, Anthony R. Martin, ed., British Interplanetary Society, 1978 (Journal of the British Interplanetary Society, Supplement, 1978), pp. S172-S179.

Ho, Darwin; and Sobon, Leon E.: Extraterrestrial Fiberglass Production Using Solar Energy. In Space Resources and Space Settlements, John Billingham, William Gilbreath, Brian O'Leary,and B. Gossett, eds. NASA SP-428, 1979, pp. 225-232.

Honda, Fujio, ed.: pp. 225-232. Methodology for Unmanned Metal Working Factory. Project Committee of Unmanned Manufacturing System Design, Bulletin of Mechanical Engineering Laboratory No. 13, Tokyo, 1974.

Honda, F.; Kimura, M.; Ozaki, S.; Tamagori, K.; Ohmi, T.; and Yoshikawa, H.: Flexible Manufacturing System Complex Provided with Laser - A National R&D Program of Japan. Proc. IFAC Symposium on Information Control Problems in Manufacturing Technology, September 1979, pp. 7-11.

IAF Conference: Robots To Produce Robots At Fujitsu. From an English-language newspaper handout at the Conference, Tokyo, 21-28 September 1980. See also The Washington Post, 13 October 1980, p. A-38.

Johnson, P. D.: Bellavior of Refractory Oxides and Metals. Alone and in Combination, in Vacuo at High Temperatures. J. Am Ceram. Soc., vol. 33, no. 5, 1950. pp. 168-171.

Johnson, Richard D.; and Holbrow, Charles, eds.:Space Settlements: A Design Study, NASA SP-413, 1977.

Kaiser Aluminum Company: New Kaiser Super Refractories, May 1962.

Kopecky, Lubomir; and Voland, Jan: The Cast Basalt Industry In Geological Problems in Lunar Research. Harold Whipple, ed., Annals of the New York Academy of Science, vol. 123, July 16, 1965, pp. 1086-1105

Luke, Hugh D.: Automation for Productivity John Wile and Sons, New York, 1972.

Miller, Rene H.; and Smith, David B. S.: Extraterrestrial Processing and Manufacturing of Large Space Systems vols. 1-3, NASA CR-161293, September 1979.

Moegling, Frank: Robot-Controlled Mold-Making Systems. Robotics Today, Fall 1980, pp. 30-31.

Mullins, Peter J.: Robot Welding of Car Bodies. Automotive Industries, 1 February 1977. Reprinted in Industrial Robots: Volume 2 - Applications, William R. Tanner, ed., Society of Manufacturing Engineers, Dearborn, Michigan, 1979, pp. 151-152.

Nagler, H.: Feasibility, Applicability, and Cost Effectiveness of LEW of Navy Ships, Structural Components and Assemblies. Contr. N00600-76C-1370, vols. 1-2, 22 December 1976.

Nitzan, David; and Rosen, Charles A.: Programmable Industrial Automation. IEEE Trans. Comp., vol. C-25, December 1976, pp. 1259-1270.

Ohmi, T.;Kanai, M.;and Honda, F.: Research and Development of a Flexible Manufacturing System Complex Provided with Laser - A Japanese National Project. Paper presented at the Workshop on Flexible Manufacturing Systems, 11-14 September 1978, Peoria, Illinois.

O'Neill, Gerard K.: Engineering a Space Manufacturing Center. Astronautics and Aeronautics, vol. 14, October 1976, pp. 20-28,36.

O'Neill, Gerard K.; Driggers, Gerald; and O'Leary, Brian: New Routes to Manufacturing in Space. Astronautics and Aeronautics, vol. 18, October 1980, pp. 46-51.

Perkins, W. A.: Model-Based Vision System for Scenes Containing Multiple Parts. General Motors Research Laboratories Publication, GMR-2386, June 1977.

Phinney, W. C.; Criswell, D. R.; Drexler, E.; and Garmirian, J.: Lunar Resources and Their Utilization. In Space Based Manufacturing from Extraterrestrial Materials, Gerard K. O'Neill, ed., AIAA Progress in Astronautics and Aeronautics Series, vol. 57, AIAA, New York, 1977, pp.97-123.

Robson, T. D.: High-Alumina Cements and Concretes. John Wiley and Sons, Inc., New York, 1962.

Schwartz, M. M.: Metals Joining Manual. McGraw-Hill Book Co., New York, 1979.

Seko, K.; and Toda, H.: Development and Application Report in the Arc Welding and Assembly Operation by the High-Performance Robot. In Proc. 4th Intl. Symp. Industrial Robots, Tokyo, Japan, November 1974, Japan Industrial Robot Assn., Tokyo, 1974, pp.487-596.

Spotts, M. F.: Design Engineering Projects. Prentice-Hall, Inc., New Jersey, 1968.

Subramanian, R. V.; Austin, H. F.; Raff, R. A. V.; Sheldon, G.; Dailey, R. T.; and Wullenwaber, D.: Use of Basalt Rock for the Production of Mineral Fiber. Pacific Northwest Commission Annual Report, Contract #NR-3001, College of Engineering, Washington State University, Pullman, Washington, June 1975. 79 pp.

Subramanian, R. V.; and Kuang-Huah, Shu: Interfacial Bonding in Basalt Fiber-Polymer Composites. 34th Annual Technical Conference, Reinforced Plastics/ Composites Institute, Section 17C, 1979, pp. 1-10.

Tanner, W. R., ed.: Industrial Robots: Volume 2 - Applications, Society of Manufacturing Engineers, Dearborn, Michigan, 1979. 287 pp.

Vajk, J. P.; Engel, J. H.; and Shettler, J. A.: Habitat and Logistic Support Requirements for the Initiation of a Space Manufacturing Enterprise. In Space Resources and Space Settlements, John Billingham, William Gilbreath, Brian O'Leary, and B. Gossett, eds., NASA SP-428, 1979, pp. 61-83.

Yankee, Herbert W.: Manufacturing Processes. Prentice Hall, Englewood Cliffs, New Jersey, 1979. 765 pp.

Zachary, Wm. B.: A Feasibility Study on the Fabrication of Integrated Circuits and Other Electronic Components in Space Utilizing Lunar Materials. Paper presented at the 5th Princeton/AIAA/SSI Conference on Space Manufacturing, 18-21 May 1981, Princeton, NJ.