Section 4.2 - Advanced Manufacturing
The first step is to develop Advanced Manufacturing, capable of self-expanding and operating remotely. It is worthwhile to develop on Earth for it's own sake because it will be less expensive to operate and thus profitable. It is then applied to building the infrastructure on Earth for space transport, and later adapt the same manufacturing methods in space. Gaining experience on Earth first makes the transition to space production easier, and the cost savings multiply to all the later projects.
Advanced Manufacturing Concepts
Using tools to build more tools is not a new idea, in fact it is nearly as old as toolmaking. What is new is to use an optimized starter set of machines which can both produce their own parts and make parts for other machines and end products. This starter set, which is called a Seed Factory, is a set of machines because no single machine can currently work with all types of materials or perform all types of production tasks. The seed factory grows in capacity by several methods:
- Adding identical copies of the starter machines,
- Producing larger versions or extensions to the starter set, and
- Producing different machines that can do different tasks and so expand the range of possible outputs.
The factory output can be divided into internal production, which is making parts for the factory itself, and external production destined for other projects or for sale.
A seed factory will not be able to produce 100% of its own parts, especially at the start. But to the extent that it does, it reduces the start-up and later production costs of whatever final factory capacity and later projects you want to complete. For example, assume the seed factory starts at 10% the size of the final factory and initially can produce 60% of the parts for expansion, and then grows to producing 90% of of its parts at full size. Then the total investment will be about 75% lower than building the final factory directly. That is enough of a reason to develop seed factories even without any later space projects. The technique will also make space projects less expensive because there is no production capacity in space right now. Self-expansion from a seed factory and local production reduces how much needs to be brought from Earth, and thus the starting cost for any major space project.
A traditional factory brought equipment and workers together in one place because transportation and communication made it more efficient to do so. The Distributed Operation concept removes the requirement that everything be in one place, and instead optimizes locations as needed. With modern communications, the control of the machinery can be a mix of a on-site humans, remote control by humans, and computer control via automation and robotics. Since all the people don't have to be nearby, you can operate in undeveloped, hostile, or expensive locations more easily, and with less of an environmental footprint. Human operators can split and re-assign their work between locations as needed if they work remotely. Likely some machines and jobs will still end up grouped together in common locations, we are merely removing the requirement that they all have to be.
Once you learn how, you can apply the distributed operation technique to space. Space is just a particularly undeveloped, hostile, and expensive location. So when you optimize your operations you will want to minimize the on-site humans, and maximize remote control, automation, and robotics. As you develop factories, habitats, and transportation systems the optimum will shift towards having more humans on-site because they can be supported locally at lower costs. Having gained experience with the distributed technique on Earth, moving to space is not something entirely new, but rather an extension of what you already know how to do.
A self-expanding and distributed seed factory is related to, but not the same as, a replicating factory. The latter is able to make all of it's own parts (100% closure), and studies of such factories have generally considered them as fully automated and at a single location, rather than a mix of direct labor/remote control/automation and multiple locations. A growing seed factory may reach 100% closure with time, but it is not assumed to start out that way. Most studies of replication assume 100% closure from the start.
The seed factory concept is related to the idea of industrial development as it relates to less developed nations or areas within a nation, but is more general in location by including space.
Automation and robotics in general has received a lot of research and development in the last few decades, and computer processor and network speeds have increased dramatically. Less attention has been given to self-expansion and distributed operation. Recent hobbyist/low cost CNC machines have sometimes been designed with making some of their own parts in mind. As far as is known, a thorough systems analysis and optimization of the seed concept described in this section has not been done. Some related work includes:
NASA Study - An early 1980's workshop, Advanced Automation for Space Missions, , whose report was originally published in paperback but now online at wikisource  and NASA  investigated replicating systems extensively. The report has since has been essentially forgotten because (1) NASA has no mandate to improve earthly manufacturing. The focus on only space missions missed the applications on Earth. (2) The estimated control system computer requirements (Table 5.1 - 2GB memory, 35 GB storage) were far beyond what was available in 1980. (3) The reference Lunar factory concept was a complete system delivered as a unit, and estimated to mass 100 tons, rather than an incremental build-up of capabilities. This implied substantial payload delivery capacity to the Moon. In 2012 the computer requirements are quite modest, and the terrestrial uses and incremental build up make it much more feasible.
KSRM Book - A 2004 book by Freitas and Merkle, Kinematic Self-Replicating Machines, exhaustively reviewed the literature of replicating systems. Although replication is not the same as starting from a seed, the ideas are sufficiently close that it is a useful reference
Open Source Project - The Open Source Ecology  is an ongoing project to develop a "Civilization Starter Kit", with an eventual 50 machines which together could produce most of their own components and support most human needs with the addition of enough farmland and labor. As a positive, they are building and testing actual hardware, which is more than most studies of closed loop or replicating factories have done. As a negative the project has used a "shotgun" approach of a large list of machines, hoping they will perform as a set. There is no evidence of a sound analysis that these are the right set of machines, or in what order or size to build them. Rather, the founder of the OSE project needed a tractor, so that was what was built first, and a workshop and cascade of other machines has followed from that.
NASA ISRU Development - The US space agency, NASA, has an In Situ Resource Utilization (ISRU) technology development program. This is aimed at using materials and energy found in space to support their missions. As such, it involves industrial elements like propellant extraction, but it is focused purely on space missions, and not on what can be done on Earth. It also tends to look at individual items, and not at an integrated growth path. Still, the technology development is useful for the research it is doing on individual elements that could be used in a more complete system.
Since the seed factory idea is new and not fully developed yet, there are open issues which more study may resolve. These include:
- The relationship among starter kit complexity, physical scale, initial cost, and growth rates.
- The optimal approach to increasing percentage closure and improved ratio of automation to human labor.
- How terrestrial or space locations affect choice of starter kit and growth paths.
- Inclusion of single-purpose end product equipment vs more generalized equipment.
- Division of production between internal growth and output for end use or for sale.
We have started a Conceptual Design Study to begin considering these issues. It is only about 20% complete as of 15 July 2012.
Advanced Manufacturing Applications
The Advanced Manufacturing concepts can be applied anywhere, but we list some specific examples in a more or less logical progression:
Seed Factory Laboratory
Technologies like automation and the Internet are not new, and are constantly being improved. Building on existing technologies like these, a seed factory prototype lab would do several things. By designing specifically for self-expansion and distributed operation we would discover what software and hardware elements are missing or would need to be changed. Actually building and operating a prototype would produce the missing and modified elements, and then demonstrate the actual performance and benefits. Initial prototypes can be built wherever convenient. A relatively small project should be able to do the initial prototyping, on a reduced scale if necessary, as it is building on existing industrial technology.
It is not likely that an ideal "universal seed" will be found on the first try, and likely never since different locations and end products would need different starting points. The progress of technology in general will also affect the best answer over time. Instead, it is more likely a series of prototype components will emerge with continued research and development. Those components would then be assembled in different ways to make up complete factories. A collection of knowledge and experience will build up on how to design, build, and operate such systems.
After building initial prototypes in an easy location, more advanced prototypes can be built in difficult locations, such as the driest place on Earth, the Atacama Desert in northern Chile, and conversely in a very wet location. That can be either a high rainfall area or on the ocean. If a prototype can work in these extremes, it should work in any condition in between. Similar testing can be done for temperature, altitude, and other variables. The first prototype can be used to produce components for these more advanced versions, demonstrating the ability to grow and expand.
Once the self-expansion and distributed operation principles are demonstrated to work for remote, inhospitable, or undeveloped locations on Earth, they can be implemented on a larger scale in numbers and size. Possible locations include:
- Hot and Cold Deserts - Deserts are defined by lack of precipitation, and besides hot deserts there are many regions where it is too cold for water vapor to produce much snow or rain.
- Oceans - Oceans lack fresh water and a solid place to build, and cover most of the Earth's surface. Deserts and oceans combined represent about 80% of the Earth's surface.
- Underdeveloped Areas - Of the remaining 20% of the planet, a significant portion is simply undeveloped, so bringing self-expanding systems there can help improve conditions there. Systems that don't require as much infrastructure, population, or outside supplies can be designed to be more ecologically sound, and jump over intermediate development stages. Thus the vast majority of the Earth's surface can potentially benefit from these techniques.
As mentioned earlier, space is a good location to apply these principles, but we will leave the details for later steps in the overall combined system. If the goal of using them later for space is kept in mind while doing the prototypes and terrestrial versions, the transition will be easier. For example, a remote control system should allow for a long time delay in communicating, since distances in space will create such delays.
For design and optimization purposes we define several type of measures to help compare different advanced manufacturing methods:
The total industrial capacity on Earth, when taken as a whole, makes all its own parts. Indeed, it built itself up, with the input of human labor, from nothing. But it is not feasible to build a starter factory with every possible machine, or to launch all of Earth's industry into space.. Thus the questions are:
- What is a minimal starter kit which can produce the greatest amount of its own parts and surplus external production?
- What component machines let it bootstrap itself to a larger and more diverse capacity with the least amount of additional external inputs?
- In what order to you add new equipment and capabilities?
There is no general requirement to package a seed factory as a single item delivered as a unit which is never added to. It will likely make more sense to start with some basic capability as a starter kit, and add to it in stages which will extend the range of outputs. These expansion kits would provide parts that can't be made internally. They would be combined with internal production to build new machines or capabilities. The combination of of starter kit and expansion kits constitutes thr total seed factory, but you can delay the later kits until you actually need them. In concept this is like a toolbox or workshop, to which you add more tools over time.
All manufacturing strives to be efficient, but typical measures of efficiency, such as useful output divided by energy input, are not sufficient of our purposes. That measure only looks at efficiency for a particular process in isolation. For a complete factory or system, we can use waste heat from a furnace, for example to heat the building or dry materials. Only energy with such a high entropy you cannot extract any more work from it is wasted at the system level.
Besides energy, we must also look at mass flow. Waste materials are those which cannot be used for some other purpose. Metal chips generated by a machine tool, for example, which can be processed back into stock for later production, are not waste in this sense. True waste is materials which cannot be reprocessed or are leaked to the environment. Also to be counted in mass flow are unique external inputs, such as carbide cutting bits for a machine tool, if you cannot make your own. The mass efficiency measure is then (useful output mass) / (total mass flow).
Productivity is an efficiency measure of the form (output) / (amount of human labor or capital equipment used). Global system productivity converts capital items into the equivalent labor to produce and assemble them. This makes sense for manufacturing systems where you make much of your own equipment and labor is the main external input. Then productivity is measured as (total output) / (direct production labor + equivalent capital item labor).
We don't know in advance all the future needs for factory growth and types of outputs. Therefore we prefer a design which has the ability to change, rather than a single fixed design. There are several ways to accomplish this:
Modularity - If we design a single machine for a given factory output, and the needs change, either up or down, we end up with wasted capacity or the need to get another machine which will have some wasted capacity. Dividing the output into smaller modular units lets you start with less total capital and then expand or contract the output more closely to the actual need. For some operations, larger scale is more efficient when running, so the amount of modularity needs to be balanced against efficiency to get the best size.
Standard Interfaces and Protocols - As paradoxical as it may sound, a standard interface can make things more flexible. Electrical outlets are standardized, but they allow plugging in almost any device at any location. Besides utility connections, standard physical dimensions and a factory dimensional grid allows placing items in any location and they will still fit. For example, if the grid unit is 1 meter, then machines, storage, and aisles are assigned space based on multiples of 1 meter. To enable dealing with different size objects, the grid can have spacings at approximate powers of two (1, 1.5, 2, 3, 4, etc) which can still fit together without wasted space. With standard software and communications, the same equipment can be operated remotely or by a local human operator interchangeably, and a changing mix of equipment can be controlled by the same software.
There are already standards for connecting automated equipment, such as the Common Industrial Protocol, and for exchanging design data, such as STEP (ISO 10303 ), so the standards do not have to be developed from scratch. A full set of standards would include physical as well as data items. Physical standards include placement and types of utilities (power, data, water, etc.) so each machine can be "plugged in" without custom design. By comparison, the PCI standards for desktop computer expansion slots are this sort of modular system. The physical, power, and data connectors are standardized so any expansion card can fit any slot of matching type.
Automation and Robotics - Factory automation is a well known technology, but usually it means using automated machines and robots to make an end product, with some amount of human labor to assist. Here we envision a more advanced version that considers the factory itself as part of the product. Full 100% automation is still not possible, so some manual remote control or direct labor by humans will be needed, but this can be reduced over time. These type of machines are programmable, so you can choose to make a single part or a whole production run with a simple software change. Thus they are inherently flexible. If the setting up of factory equipment locations, storage, and other items can itself be automated, then the entire factory can become configurable and flexible according to changing needs. Automation and robotics can also improve productivity, both by being faster than humans, and being able to operate almost 100% of the time without tiring.
Distribution Network - Given sufficient bandwidth, the equipment can be directly controlled by remote humans as an alternative. For example, you may want to place a solar panel factory in the Sahara Desert, because it has abundant sand for raw materials and sunlight for power. With remote operation your staff does not have to live in the desert if they don't want to. Remote operation is currently used for military drones, deep sea vehicles, spacecraft, and some types of mining, where the environment is hazardous or expensive to support humans. With recent improvements in electronics and network bandwidth it can be effectively applied to more tasks than before. For future space projects, remote operation has obvious benefits extending your operational reach, from where you currently are to new locations.
Making your own parts can be called “closing the loop” in the sense that the output of parts loops back as an input to making a copy or expansion to the factory. You can measure closure as a ratio of (self made mass)/(total mass needed). Another measure is return on mass, which is (total ouput)/(factory mass). As an alternative to measuring by mass, you can measure closure on cost.
Analyzing closure ratios is a stepwise process working backwards from the end products. You first identify which machines and processes you need to make the end products. From that you can identify which equipment you do not already have in place. For the missing ones you can further determine how much of those you can make internally. Eventually you trace everything back to a machine or material you have or can make, or to those you can't. The ratio of internal make to end output is your closure ratio. In doing such an analysis, what would otherwise be a waste product from one process should be considered for recycling into another process.
If you try to reach 100% closure, in theory you will reach some limit of starting machines that can make all the others including themselves. We know our entire industrial civilization can do this, so some smaller subset of at least one machine of each type should also be able to also. In practice, a few processes, like making computer chips, are difficult and expensive to do in small quantity. For those it will usually not make economic sense to make your own. The few previous studies on this kind of closed loop production found around 2% of the total items were not practical to self-make, or in other words 98% closure. Still, having to buy or import 2% is a great improvement over 100%.
Designing a distributed seed factory will likely never be finished in the same sense building a home workshop or manufacturing business is never done. Rather, you have some starting point which is productive in itself, to which you add more capability over time. Also, experience will teach you better ways of doing things, and technology in the rest of the world keeps improving. So instead of trying to get it all working at once, it makes sense to start with something simple, and design for expansion and upgrade. This is similar to the incremental development of software, and in fact a large part of a networked and automated factory will be software.
The starting point should aim to produce an increasing amount of it's own parts in terms of mass and cost. For example, on Earth, if you start with cement making, it being the most expensive component of concrete, you can supply a large part of your own buildings by weight, and cut cost dramatically. Then if you add steel fabrication, you would add another large segment of the total building, and so on. The design should also take into account flexibility. Rather than mass production of a single item, this type of factory will likely mix producing items for sale and making new parts for itself, depending on the schedule of work orders.
Prior to doing the design and prototypes, we can guess a terrestrial starter kit will include machines for processing wood, basic metals, ceramics including clay and cement for concrete, rock, and possibly glass and plastics. In addition, some kind of energy supply, for example based on sunlight or biomass, would be needed, and computers and communications to tie it all together. This guess is based on those being the basic materials from which most things on Earth are made. For the different conditions in space, a different starter kit will almost certainly be needed.
There are several approaches to actually design a starting point. One is to simply assume a starter kit, like the one in the previous paragraph, and analyze it to determine if items need to be added or removed. Another approach is to start with the machine list of a well equipped machine shop (see Appendix 2), since machine shops can generally make parts for any other kind of machine. From the starting point, we can then look at alternatives, optimizations, and upgrades.
For space the starting point will be different because of the different environment. The advanced manufacturing discussion for those will be left to later sections of this combined system example: Orbital Assembly and Processing Factory.
The same engineering process is used to go from an assumed starting point to a final seed factory design as for any other engineering design. The general process is described in the Systems Engineering and following sections. It involves tasks like breaking up a complex design into simpler parts which together make up the full system. The simpler parts are then easier to design and optimize. While the various engineering methods are used to evolve the design, it will not be "finished" in the traditional sense of completing design and then going into production. Once it reaches a good enough state, a prototype or first production version can be built, but further work and feedback from existing versions will lead to improvements over time. This is more like the common software development model where a version 1.0 leads to later versions over time of the same basic program.
In addition to hardware components, a suite of software tools will be needed. These will not change that much even as the hardware parts vary dramatically. Some of them are existing software, others will need to be customized or written from scratch:
- Computer Aided Engineering (CAE) - Software to generate 3D models of a design, then analyze it for stress, temperature distribution, and other features. A number of such programs exist.
- Matter Compiler - Software to convert CAE models to low level instructions for a target set of machine tools and robots. This is similar to how a software compiler turns high level program instructions into machine code for a computer processor. This is not known to exist at the whole factory level of machines, although converters exist for individual machines.
- Tool Drivers - These are the interface between the computer and the individual machines in a factory, so that commands and files can be sent to operate the machines.
- Augmented Reality Simulator - To provide a virtual factory or construction environment overlaid on real views to try out remote control methods and factory procedures. It is much less expensive to try things out in a simulator than to "bend metal" and find out the instructions are wrong.
- Remote Operations Software - For humans to directly control robots and automated equipment at a remote location. This should use the same interface as the simulator. Some tasks will not be able to run automatically, so between this software and live humans at the factory the non-automated tasks will get done.
Example Seed Factory Designs
We look at several example designs on Earth to see how the choice of location and environment affects the starter kit.
Temperate Prototype Factory
The temperate location prototype assumes typical conditions in an already developed and populated area. This included moderate rainfall and temperature, typical soils, and availability of public infrastructure. A minimal starter kit might consist of a storage area for parts and materials, a general purpose CNC machine tool, and a robot to transport items from place to place and assemble parts using multiple tools attached to the robot arm, all controlled remotely by humans or by pre-planned manufacturing files. A more comprehensive one will have more machine tools, robot variations, and the ability to work outdoors.
CNC Machine Tools
This is a set of modular components that can be assembled in various configurations for different tasks. It includes one or more of the following:
- Fixed solar array charging station
- Replaceable battery packs - can be left at the charging station and swapped as needed, or charge the robot by parking it.
- Base chassis in various sizes to which other items are attached
- Wheel kits for indoor factory or outdoor use
- Seat kit for passenger use
- Camera and control unit for remote use
- Robot arm with various end tools stored in a toolbox
- Base plate for non-mobile robot
- Pallet/Box attachments for cargo
- Other power attachments for materials handling, stabilizers, excavator arms, etc.
Ocean Prototype Factory
Alpine Desert Prototype Factory
- Robert Freitas, ed., "Advanced Automation for Space Missions", NASA Conference Publication 2255, NASA, 1982.
- Advanced Automation for Space Missions at Wikisource
- Advanced Automation for Space Missions PDF version at NASA
- Open Source Ecology Wiki Main Page