Unmanned Systems E-Mag 2016 by OpenSystems Media

UNMANNED Systems

E M A G

2016 Volume 3 Number 1

ññ Replicating human reasoning

with sensor-driven mobile supercomputing

ññ Powerful enabler of modern

military systems

ññ Prototyping for defense tech via

additive manufacturing reduces time costs

mil-embedded.com/topics/unmanned-systems

Sponsored by: Abaco, Kontron, Stratasys, Annapolis Micro Systems, Sealevel

CAPABILITY G IN G A IM R E S ALL-WEATH IT MAY BE IT RET IONARY TUR T U L O V E R S IT MAY BE IT N 0MM CANNO 4 S IT E B IT Y A IT M Y POWER UN R IA IL X U A T S ULTRA-QUIE e most IT MAY BE IT

o th g vehicle is als n ti h g fi d re o arm ired ost advanced f sensor-acqu o s B T 6 re o The world’s m st and en apture, process it Ethernet op b a ig G /s b G 0 digital. It can c ver a 2 alable ansmitting it o -of-the-art, sc tr , te n a o st ti a a h rm it w fo d in can be uippe ckbone. It is eq t which targets a a b e g re n tu ra c e e it th h s arc ouble at more than d th e g a k c a p R A IST tracked. identified and ven. Not only ro p t u b , d e c n a e. only adv ut for its lifetim deploys is not b it y, a y d g lo to o r n fo h c ly The te d not on ly reliable. An d e g g ru t u b , reliable IGHTING

RMORED F A D E C N A V D MOST A Y THE WORLD’S ION READY B S IS M E D A M LSO S VEHICLE IS A ACO SYSTEM B A M O R F Y TECHNOLOG

Watch Video

ABACO SYSTEMS: EMBEDDED IN THE WORLD’S MOST CRITICAL MISSIONS SINCE 1986.

COMMUNICATE WITH CONFIDENCE COTS and Full Custom Military Solutions The modern battlefield demands ruggedized equipment that deliver a combat advantage. Sealevel computing and communications solutions are proven effective even in the harshest environments. Choose from an array of COTS products or partner with us to create a custom solution ideally suited to your particular application.

Watch Video

SYNCHRONOUS COMMUNICATIONS

POWERFUL SOLID-STATE PROCESSING

The Relio R2 Sync Server solid-state computer offers designers robust synchronous serial communications, small size, powerful processing, and long product life. Perfect for radar, satellite, and other military applications, the R2 Sync Server's four synchronous ports are individually configurable for RS-232, RS-422, RS-485, RS-530, RS-530A or V.35. The system operates over a wide 0-50Â°C temperature range and is compliant to MIL-STD-810 specifications for shock and vibration.

sealevel-mil.com â&#x20AC;˘ 864.843.4343

Photo Credit: US Air Force

UNMANNED Systems

E M A G

2016 Volume 3 Number 1

Featuring

Replicating human reasoning with sensor-driven mobile supercomputing By Abaco Systems

Powerful enabler of modern military systems By Kontron

Prototyping for defense tech via additive manufacturing reduces time costs By Stratasys

Sponsored By

ÂŠ 2016 OpenSystems Media, ÂŠ Military Embedded Systems. All registered brands and trademarks within the Unmanned Systems E-mag are the property of their respective owners.

SKY-HIGH PERFORMANCE THAT’S BUILT FOR ABUSE

& √ UltraScale™ & UltraScale+™ ® ® √ Arria 10 & Stratix 10

WILD40 ™ 40GB 6U & 3U OpenVPX EcoSystem

MADE IN USA

All Systems Include Open Project Builder ™ Our Vendor-Independent FPGA Development Tool See a demo at www.AnnapMicro.com/OPB

Replicating human reasoning with sensor-driven mobile supercomputing Object classification using NVIDIAâ&#x20AC;&#x2122;s Advanced Driver Assistance System. Photo courtesy of NVIDIA.

The human brain remains the most effective classifier on earth, able to perform real time differentiation between a huge number of disparate objects and events, in varying environmental conditions of lighting and occlusion.This ability is learned from birth, and while there is significant overlap of experience leading to the illusion that we all classify in the same way, each human is in fact operating in a highly individualistic manner. In fact, the classification is actually part of a wider cognitive process in humans and so outlier objects, behaviors, and different individuals depending on their prior experiences can interpret events differently. Furthermore, this interpretation is shown to be affected by many factors including the emotional state of the observer, and sensory loading. This is most notable in the evidence given by multiple witnesses to events that fall outside the â&#x20AC;&#x153;normalâ&#x20AC;? human experience, such as major accidents or crimes; in that moment, each human cognitive process has been highly focused on processing the new and

unusual events, meaning that normally reliable functions such as memory or behavioral reactions have become secondary, leading to questionable subjective recollection. Efforts to replicate the human ability to classify have historically been largely academic, lab-bound and isolated; each employs one chosen algorithm, is often coded, tuned, and operated on the same platform; and does not apply continuous learning or evolution. In the world of the Industrial Internet, there is an opportunity to have many

different machines observing the physical world through a variety of sensors and applying a “collective learning,” such that the experience of the individual machine contributes to the learning of the whole. In this model, the processing capability exists in large data centers at the heart of the network; here, there are sufficient computing resources to tackle the problem of algorithm tuning and continuous learning. The edge nodes of the network interact with the real world sensors and apply processing capability to the problem at hand, classifying, filtering, and reducing the vast quantity of sensor data into a useful set of information to pass to the rest of the network, as well as identifying errors and failures in the algorithms. The datasets associated with these failures can be transmitted to the central resource for analysis and further learning to improve the algorithms (and hence performance) at the edge nodes. The benefit of the model outlined above is that the learning experience is not confined to a single machine, but is distributed for the whole network to benefit from.

Current state of the art There have been many years of research into video and image processing, with a great many mathematical strategies proposed to solve a variety of problems. Early research was hindered by the lack of computational resources, requiring huge and expensive machines to run the coded algorithms on stored video clips and image stills. As Moore’s law predicted, the ever-increasing density of integrated

components means that processing performance gets greater each year; the capability of ICs designed for the mobile industry now puts colossal computational capability into a size, weight, and power (SWaP) envelope that ensures video and image processing will soon be ubiquitous in a wide variety of applications. One example of such an application is the work undertaken by NVIDIA to implement advanced driver assistance systems (ADAS). Here, the edge nodes of the system are high-performance System-on-Chip (SoC) processors mounted inside individual cars. The SoC runs several different types of algorithms to classify objects and also determine something about their motion, pedestrian versus car, stationary car versus moving car. The objects and situations that cause the classifier to fail will be uploaded to the cloud for analysis on huge clusters, and the resulting improved classifiers sent to all cars in the network. The result is that every car learns simultaneously from the experience of an individual. The network is objective in its response to a particular object or event, regardless of the situation. (See lead image). The latest NVIDIA SoC, the Tegra X1, is a “mobile supercomputer” comprising a quad-core ARM architecture coupled with 256 graphics processing cores based on the “Maxwell” architecture. This combination makes the processor extremely effective at handling visionbased tasks based on parallel processing of the large datasets produced by video cameras. The NVIDIA Tegra X1 is rated at 1 TeraFLOPS of processing performance, with a single computer module

measuring the size of a credit card and consuming less than 10 Watts implementing this SoC. This is remarkable, given the fact that the first computer to reach 1 TeraFLOPS of processing performance was the ASCI Red supercomputer built by Intel and Sandia National Laboratories in 1996; it measured 1,600 sq ft and consumed 850 kiloWatts. ASCI Red remained the most powerful computer in the world until 2000. In the not too distant future, every car on our road is likely to carry several TeraFLOPS of processing performance!

Classification algorithms Support Vector Machine (SVM) is a “supervised learning” model, in which the system is given training examples that fall into a category and others that do not. The algorithm maps the examples into a multi-dimensional space, and then attempts to find a plane through that space with the greatest separation between those in the category and those not. New examples are then mapped into the space and classified according to which side of the line they fall. The SVM algorithm dates back to the 1960s, but research has led to progressive improvements in it, including dealing with misclassification and nonlinear solutions. However, the quality of the solution depends on the quality of the initial example data set (Figure 1). An alternative to SVM is the Cascaded Haar algorithm, which is particularly suitable for classification of objects within an image. The system is again trained using a set of example images. A series of classifiers detecting edges, lines and center-surround features are

Figure 1 | SVM classification. H1 has a limited distance to the two groups; H2 achieves maximum distance to the two groups; H3 does not separate the two groups. scanned across the image, returning a positive or negative result if the feature matches the image. The algorithm applies multiple classifiers to the image until an object is either recognized or rejected. The classifiers may be grouped into complex classifiers via a number of weighting techniques. This cascade of classifiers is then applied to the candidate image until all the classifiers are either passed or rejected (Figure 2). There is currently much implementation effort being applied to machine learning using deep learning techniques; sophisticated, multi-level “deep” neural networks (DNN) optimized for GPU. This technique uses nodes, or “neurons,” connected together to mimic the biological neural network found in the human brain. During learning, the connections develop adaptive weightings to simulate the strength of connections between neurons. Layers of neurons connected together translate their input via the

weighting to the next layer. The more complex the system, the more layers that need to be implemented and the more processing performance that is required; hence, various techniques exist for optimizing the connectivity. DNN algorithms are very well suited to object classification for images and video (Figure 3). The neural network may be trained to recognize many objects under different 1. Edge features

(a)

(b)

(c)

(d)

2. Line features (a)

(b)

(c)

(d)

(e)

(f)

conditions, but due to the way in which the weightings must propagate both forwards and backwards through the network, the learning phase is very computationally intensive. In deployment, the network only propagates in the forward direction, and so can be run much more efficiently on small, deployable hardware. In addition to these object classification techniques, interframe temporal processing allows for added information about the objects to be derived; are they static within the scene, or are they moving, and with what velocity relative to the viewer? (g) (h)

3. Center-surround features

(a)

(b)

Figure 2 | Haar-like feature classifiers.

Figure 3 | Schematic representation of a deep neural network, showing how more complex features are captured in deeper layers. Photo courtesy of NVIDIA.

Implementation relies on high performance hardware The implementation of object classifiers on deployable hardware relies on high performance hardware and very well optimized libraries. Abaco Systems has selected the NVIDIA Tegra System-on-Chip as the basis for a new product because it meets both of these criteria. The first rugged product, the mCOM10-K1, is a credit card-sized

Figure 4 | Measuring 84x55mm, Abaco’s TEGRA K1 computer on module. computer-on-module using the Tegra K1, with quad-core ARM A15 CPU and 192 “Kepler”-architecture GPU cores, delivering 327 GFLOPs in a 10 Watt power envelope. Future developments will incorporate the recently announced Tegra X1, driving up the performance and increasing the functional capability. Meanwhile, NVIDIA is continuing to invest in the development of highly optimized libraries to support object classification, including the OpenCV Open Source Computer Vision library, the cuDNN CUDA Deep Neural Network library, and the underlying CUDA itself. Abaco draws on both the SoC and the optimized libraries to create compelling products and demonstrations of this technology, and a roadmap to ensure that investments in technology today are protected into the future (Figure 4).

Future concepts Along with an ongoing technology insertion roadmap, Abaco is integrating these computer vision functions with other key technologies into a broad range of advanced applications that

will be transformative to our lives. For example, a computer-based classification system linked to a Data Distribution Service (DDS) or OLE for Process Control (OPC-UA) networking middleware will allow for real time exchange of data between different edge nodes within a larger community of computers. The individual pixels are not needed in the larger system network in order for the information about the detected objects to be disseminated and correlated, even when observed from different viewpoints. Assuming video can be stored on individual vehicles, it would also be possible to interrogate each edge node for retrieval of video clips recorded at a geo-specific location before and after a specific trigger, allowing for a comprehensive review of events leading up to and following an accident. As learning algorithms break free from the confines of the computer rooms of research institutes and into the real world, the evolution of these algorithms is expected to progress rapidly, allowing classifiers to distinguish between an ambulance and a delivery truck, or a school bus and a van, for example. The roadmap of low SWaP computers with colossal processing performance is set to keep pace with the demands from the researchers’ algorithms and propel us beyond object classification into complex behavioral classification. The ADAS systems of the future will all objectively understand that a small human is more likely to run into the road than a large human, and that probability is increased if the small human also has a ball…  www.abaco.com

Powerful enabler of modern military systems

Small form factor design has become a powerful enabler in the modern military, allowing defense system developers to meet increasing integration requirements for a growing group of unmanned and portable systems. Going a step further are today’s pre-validated small form factor systems that represent the next generation in these designs, ensuring flexibility and creating a trusted commercial-off-theshelf (COTS) platform that reduces development cycles and accelerates Proofof-Concept (PoC) development. Kontron’s COBALT is a prime example of a rugged, pre-tested platform that leverages the inherent flexibility and advantages of Computer-on-Modules (COMs). Proven in a broad range of successful deployments, Kontron’s next generation high-performance embedded COBALT computer provides a powerful and valuable standards-based design resource optimized for the rigors of mil-aero computing.

Unmanned and portable systems are modern military essentials Rapid changes on the front line are driving demand for embedded computing solutions that deliver high performance, reduced size-weight-power in an environmentally ruggedized platform in order to meet a diverse list of portable, unmanned, and robotic system requirements. According to a 2014 report issued by the Center for Strategic and International Studies (CSIS), in 2013 the Department of Defense (DoD) possessed nearly 11,000 unmanned aerial systems (UAS) representing a wide variety of sizes and capabilities. Technical assets also included thousands of small unmanned ground vehicles (UGVs), and an assortment of experimental unmanned surface vessels (USV) and underwater vehicles. The strategy for deploying these systems is to handle jobs that are considered ‘dull, dirty, and dangerous,’ or for life saving purposes countering hybrid threats such as organized militias, improvised explosive devices (IEDs), or uncertainty about location, context, and duration of adversarial contact. Modular plug-and-play payloads maximize safety and combat capability in these scenarios – improving situational awareness that protects soldiers and empowers enhanced leadership decisions. Even with defense budget cutbacks, the OEM opportunity is significant. The DoD has stated that battlefield commanders have a nearly insatiable need for persistent intelligence, surveillance, and reconnaissance (ISR), which must be ideally suited to the spectrum of high-performance unmanned systems.

For example, Air Force One resources have shifted to focus on MQ-9 procurement, with 600 percent more payload capability than MQ-1. Combined with the versatile and powerful Wide Area Airborne Surveillance (WAAS) sensor, the MQ-9 Reaper is increasing the effectiveness of individual Combat Air Patrols (CAPs) by more than 1,200 percent initially, and will eventually reach a 6,000 percent improvement over today’s MQ-1 Predator sUAS. No longer is the military concentrating on the particular platform such as a truck or other vehicle. The new focus is the ‘integrated system’ itself, and includes the network and payload along with processing, exploitation, and dissemination (PED) capabilities. Literally hundreds of systems may be fitted onto a single aircraft or ground vehicle in order to meet these needs, and COMs are just one of the key standard technologies that respond with ideal flexibility, efficiency, and value.

Capitalizing on a trusted COTS platform and a costeffective, flexible design Operating as a complete computer mounted on a carrier board, the COM Express speciﬁcation offers one of the smallest standards-based form factors available for military systems; I/O is customizable and the system allows a range of module sizes that optimize size, weight, and power (SWaP) considerations. Integrating mezzanine options adds further value, allowing developers to create new systems without significant modification to an original base design. This

illustrates the building block concept that simplifies development. Packaging COM Express-based COMs in its ruggedized housing, COBALT creates an application-ready platform ideal for rugged, high-performance deployments. The COM processor board forms the heart of COBALT’s small form factor system, and the complete system includes rugged baseboard, power module, housing, and appropriate I/O connectors. The Kontron COBALT is fanless and fully enclosed; the system weighs less than six pounds and offers efficient thermal management in a small 8.5 (W) x 5.5 (D) x 3.9 (H)-inch form factor. Scalable processor options allow developers to manage computing

performance as needed for specific application requirements. For instance, the system can perform based on very low power Intel Atom processors or be deployed as a more powerful Intel Core i7-based system. These modules include a range of powerful hardware-based security and performance features such as Intel Trusted Execution Technology (TXT), Intel Virtualization Technology (VT), and Intel Active Management Technology (AMT). Intel TXT is built into the silicon and helps close any security gaps by evaluating the launch environment and enforcing only ‘known good’ code execution; Intel VT provides a hardware assist to virtualization software, reducing

Figure 1* | COBALT is a highly scalable embedded computer system, sealed and validated IP67, available with a wide selection of processor, storage, power, and interface options. Its small footprint and low power are optimized for applications requiring both performance and reduced SWAP.

its size, cost, and complexity, and paying special attention to reducing virtualization overheads occurring in cache, I/O, and memory. Intel AMT adds efficiency by enabling end-users to query, restore, upgrade, and protect devices remotely. For military OEMs and developers, this means that devices and systems can be diagnosed and repaired remotely – providing a competitive edge and helping meet military demands for effective performance and guaranteed uptime for mission-critical systems. Strict tolerance to shock and vibration has been pre-validated based on the broad spectrum of UAV, tracked vehicle, shipboard, and avionics environments. COBALT is also tested to perform reliably in temperatures ranging from -40° to +71° C, proven ready to handle the environmental and physical rigors of military applications. The result is a very robust, standards-based platform that delivers the performance, size, weight, and flexibility designers need to streamline development of many different types of applications – all from a single platform that enables the use and reuse of low power, reliable designs.

Feature-rich, pre-validated platforms accelerate proof-ofconcept COBALT also provides a perfect-fit design approach via use of the system’s carrier board and mezzanine card options. Additional design value is also achieved through COMs inherent processor upgradability. The result is a

fully tested, optimized and cost-effective design path that reduces time-tomarket and satisfies an extensive range of mil-aero design requirements. These benefits also help defense contractors to quickly meet today’s PoC and prototype design needs by adding their own application-specific software. No extensive costs or extended development timelines are required in order to test system functionality. Design processes are streamlined and deployment is faster when developers can focus resources on the I/O customization aspects of their system, including support for both Linux and Windows operating systems. Flexible I/O can be designed in through application- specific customization on the carrier board. COBALT COMs’ standard features include LAN, SATA, video, audio, GPIO, configurable serial ports, and multiple USB. Other unique features enhance rugged performance, such as a special Rapid Shutdown circuit design on the system’s RXT modules. Rapid Shutdown is an onboard mechanism that quickly powers off the system, preventing damage in an emergency and allowing the system to survive a high-energy pulse such as a nuclear event or high-energy electromagnetic pulse (EMP).

Forward thinking with Type 6 COM Express COBALT also leverages the COM Express Type 6 pin-out and its future design options as a means of ensuring longevity to system designs. Legacy PCI pins (previously assigned to the IDE interface in the Type 2 pin-out) are real-located to

enable the digital display interface, additional PCI Express lanes, or technologies still in development. For instance, future technologies could include Super-Speed USB, with 16 free pins offering sufficient lines to implement four of the eight USB 2.0 ports as USB 3.0 ports instead. Type 6 also offers configurable Digital Display Interfaces (DDI) SDVO, DisplayPort, and HDMI/DVI along with 23 PCI Express Gen 2 lanes.

performance jump (compared to devices incorporating earlier pin-out options) increases the versatility of the system and establishes an enhanced fourth-generation graphics architecture. These performance values are ideal for unmanned applications such as high-resolution surveillance for situational awareness.

“ COBALT CAPITALIZES ON CLOSE MANUFACTURER RELATIONSHIPS WITH MAJOR SUPPLIERS SUCH AS INTEL.”

Persistent surveillance or high-resolution imaging applications illustrate the potential of improved graphics performance. More powerful graphics display and processing features allow military professionals to simultaneously access and process multiple displays of information. Informational data can now be reviewed in the field or stored for review, while urgent information can be quickly distributed for immediate action. Further, the addition of native support for all the newest display interfaces simplifies carrier board designs. This is significant in reducing time-tomarket and total cost of ownership for graphics-intensive military applications. The extensive PCIe support in Type 6 underscores the overall trend of moving away from legacy parallel interfaces and towards pure serial embedded system designs – a perfect fit for military applications that demand higher bandwidth and reduced latency.

Type 6 gives designers more to work with than in Type 2, such as greater native display options and higher serial bandwidth. The real advantage is that some of the extra PCI Express lanes can be routed to serial-based mezzanine card slots such as mPCIe and XMC to create expansion options. The resulting

Managing costs while adding application-specific software

In such a design scenario, ruggedized, thermal performance is validated at the board level and includes reliable performance in temperatures ranging from -40° to +85°C. All I/O routed

from the baseboard relies on a proven rugged connector, while all external I/O employs a 38999 type MIL circular connector. COBALT ensures the system is modular and upgradable, evolving performance by swapping out modules to access processor advancements. Proven designs can be reused in smaller systems, giving OEMs a competitive advantage in meeting design requirements while avoiding additional customization costs and development resources. Developers avoid design requalification or additional PoC requirements because the system’s baseboard stack provides all necessary interconnects between the COM Express board and XMC and mPCIe interfaces.

Preparing for increased performance in the same small space Portable or unmanned systems may include unmanned aerial vehicles (UAVs), vetronics, robotics, and avionics systems. Each of these computing devices or environments allows no option to expand system footprint in order to accommodate increased power or performance. The need for upgrades is inevitable, as demand for ever-increasing sensor capabilities continue. Yet, it is not practical to simply add more equipment onto an unmanned vehicle. Together these challenges increase the necessity for a compact and modular-open-standard high-performance module or system, making COBALT an ideal computing small form factor solution.

Optimized as pre-validated military design platforms, the Kontron COBALT is shown to reduce development cycles, enable fast PoC and act as a versatile, standards-based building block for the range of defense system deployments. Furthermore, COBALT COMs’ ability to couple mezzanine modules with carrier boards enables military designers to readily use and reuse low power, reliable designs in the development of smaller, highly rugged, high performance systems. COBALT capitalizes on close manufacturer relationships with major suppliers such as Intel. In addition to the inherent flexibility to select a range of processors, these relationships assure developers that components are sourced from long-term embedded roadmaps eliminating the risk of unplanned design changes and unexpected application modification. COBALT is the epitome of the ideal building block approach defense contractors can now use to accelerate development, helping them meet today’s small form factor requirements and keeping future systems poised for what’s next to deliver advanced defense system capabilities. Figure 1* // According to a report from the Aerospace Industries Association, UAV platforms are considered assets in providing critical capabilities to the U.S. military including Persistent Intelligence, Surveillance and Reconnaissance and improved accuracy for tactical strikes. Size and capabilities vary dramatically, ranging from the “micro” U.S. Marine Corps Wasp, weighing just a few pounds with a onehour flight endurance, to the Air Force RQ-4 Global Hawk, weighing 7,600 pounds and capable of remaining airborne for 32 continuous hours at altitudes as high as 60,000 feet. See page 7 at: http://www.aia-aerospace.org/assets/ AIA_UAS_Report_small.pdf

 www.kontron.com

Prototyping for defense tech via additive manufacturing reduces time costs

Engineers can streamline design of unmanned aerial vehicles thanks to 3-D printing.

Developing any sort of defense application, whether unmanned aircraft or shipboard weapons systems, requires building prototypes of components that can be tested and iterated until they achieve maximum safety and efficiency. The components and systems involved must not fail; otherwise, human lives could be lost. It can be expensive, however, to build prototypes, because they are often built of the same materials as the deployed parts. This is no longer necessarily true, though, thanks to the advent of additive manufacturing. Defense and aerospace engineers

can now produce parts through 3-D printing, significantly reducing the time and cost of prototyping.

The challenges of prototyping Building and testing parts is a necessary but often time-consuming part of system development. An aerospace engineer might, for instance, have to test a component of an aerial vehicle, alter the design slightly, test it again, and keep doing that until he or she has achieved the desired performance. That sort of iteration can be expensive when the components are machined out of production quality materials, and repeating the steps adds to development time. Depending on the complexity of the design, building these parts can cost from several hundred to several thousand dollars, and the time between new versions can be measured in weeks.

of material used - another cost savings. There can also be less need for post-processing, with less wasted material and quicker production.

Faster and cheaper UTC Aerospace Systems builds critical assemblies for both commercial and military aircraft, which requires the company to operate a well-equipped machine shop. The company purchased a Fortus 900mc 3-D Production System from Stratasys so it could build prototypes more quickly. In one case, technicians wanted a redesign of a nozzle for a fume collection system used with a machine tool. Using ULTEM 1010 resin, a thermoplastic with high tensile strength, heat resistance, and chemical resistance, they built the nozzle. Total cost fort the part was $750, and production took one day. Producing the same part through CNC machining would

3-D printing, on the other hand, can cut the cost by at least two-thirds over machined metal parts, and production that took days may take only hours. This reduction in turnaround time also saves money by speeding up the development process, Faster product launch times reported by Stratasys customers* and it allows engineers to quickly incorporate new ideas into % o their designs. Additionally, the f r materials used by 3-D printers, 29% e s 25% 25% including thermoplastics, can p o have the necessary mechanical n d 14% e properties, such as strength, heat n t resistance and flexibility, for con7% s ducting tests. ÂŽ

The differences between additive manufacturing and traditional techniques bring added advantages. Because parts are built up in layers, old considerations for tooling paths no longer apply, which can reduce the amount

1-5%

5-10%

10-25%

25% or more

Time-to-market savings *Based on a survey of over 1,000 Stratasys 3D Printer owners

Figure 1 | Stratasys customers saw a reduction in time to market through the use of 3-D printers.

have cost $2,000 and taken 21 days. 3-D printing resulted in a 63 percent reduction in cost and a 95 percent reduction in time (Figure 1). Bell Helicopter builds the Osprey, the hybrid aircraft that can lift off, hover, and land like a heliFigure 2 | This new Y-shaped design in strong ULTEM 1010 resin has been in copter but also cruise production use for eight months without cracking. at jet speeds like an airplane. Recently, Bell decided to test an upgrade to make millions of parts are usually made the tail wiring. That involved installing of steel with a CNC milling machine or branching conduits for the wires in six via electrical discharge machining. Such sections of the vertical stabilizers. After molds can cost hundreds of thousands initially testing one conduit, engineers of dollars and take months to produce. varied the design slightly and ordered In cases in which the mold only needs five more complete sets, for a total of to last through the production of tens 42 conduit models. With the Fortus of thousands of parts, soft tool molds system running round the clock, it took can be made of aluminum. While less two and a half days to make the entire expensive, these still cost from $2,500 collection. Had Bell used the traditional to $25,000 and take two to six weeks prototyping technique, using cast alu- to make. The time and expense make it minum, they would have had to wait six difficult to justify building several iteraweeks. Because the printing process tions of molds to test variations in the is so efficient, it allows the engineers parts they will produce. to create more iterations than would otherwise be possible, resulting in bet- On the other hand, 3-D printing PolyJet ter-designed components (Figure 2). technology from Stratasys allows for the creation of prototype molds. The proBreaking the mold cess involves positioning multiple layers One area where prototyping can be of a liquid photopolymer into desired prohibitively expensive is in the design configurations, then curing it with UV of molds used for producing parts light. Once cured, the molds can be through injection molding, in which placed into injection molding equipplastic is injected into a mold, then ment to create prototypes from the cools and hardens to the shape of the same material that will be used in the cavity. The molds used in production to final product. That allows engineers to

Reduced design cycle iterations by having a Stratasys 3D Printer in-house* % o f r e s p o n d e n t s

Keeping everything in-house 26%

27%

14% 6%

1-5%

5-10%

10-25%

25% or more

Reduction in iterations

*Based on a survey of over 1,000 Stratasys 3D Printer owners

Figure 3 | Having access to a 3-D printer reduces the number of design cycles engineers must go through. produce parts that are the same as the finished product, which they can use to gather real-world performance data. The PolyJet molds perform in the same way as metal molds, but are much cheaper, easier, and faster to make. Though these molds can only be used to make approximately 100 parts — depending on factors such as design complexity and the type of thermoplastic used — they can be built within a few hours, with a low average mold cost. Designers and engineers have the freedom to vary the design and quickly iterate new versions. The 3-D printing process also allows molds to be designed with complex geometries, thin walls, and fine details, which are not always achievable with traditional mold-making processes. A mold made from Acrylonitrile Butadiene Styrene can be built in 30-micron layers with accuracy as high as 0.1 mm, creating a smooth surface finish that

generally does not require post-processing, speeding up production time. Having a 3-D printer in-house can realize all the advantages of additive manufacturing. In-house printing can cost from a third to a fifth as much as outsourcing prototypes. Consumer product manufacturer, Akaishi, brought their 3-D printing in house with an FDM Dimension 3-D printer. After making the switch from outsourcing their prototypes, Akaishi estimated that having at their facility cut their costs by 73 percent (Figure 3). While outsourcing can require weeks to get a prototype part back, a printer— which can run overnight and during weekends—can reduce that time to hours. That allows designers to quickly test and redesign their parts, shaving weeks off development time. It also permits them to fine-tune the design before expensive molds and die casts are made, reducing the risk of manufacturing errors. Given the importance of creating efficient, reliable components for defense applications, prototyping plays a critical role. Designing parts and testing their efficiency is key to maintaining performance and safety, and prototyping and iterating are critical to that design process. Additive manufacturing improves prototyping by significantly reducing the time and cost of making multiple prototypes, allowing defense and aerospace engineers to print 3-D parts until they are satisfied with the final product.  www.stratasys.com

OpenSystems Media works with industry leaders to develop and publish content that educates our readers.

Check out our whiteÂ papers. http://whitepapers.opensystemsmedia.com/

Most popular topics: Managing SWaP COM Express MIL-STD-1553 Cockpit Display Systems Thermal Management Shock and Vibration Testing Radar Software Defined Radio

FPGAs COTS VPX UAVs PICMG Counterfeit parts Data Security

UNMANNED Systems