DARPA | DARPA DMACE (Digital Manufacturing Analysis, Correlation and Estimation) Challenge
  This is an archived web site made available for research purposes. The page is no longer being updated or maintained.


(Digital Manufacturing Analysis, Correlation and Estimation)

Frequently Asked Questions (FAQs)

October 29, 2010

Table of Contents:
  1. New Questions: Dec 3rd
  2. New Questions: Dec 2nd
  3. New Questions: Nov 30th
  4. New Questions: Nov 29th
  5. New Questions: Nov 18th
  6. New Questions: Nov 12th
  7. General
  8. Eligibility
  9. How the DMACE Challenge Works
  10. Prizes
  11. Download PDF version

1. New questions: December 03, 2010

Can you verify the units and dimensions of the sphere?
The outer diameter is approximately 40mm and inner diameter is approximately 30mm. We use 'approximately' since the beam velocity and current affect how much material is deposited while manufacturing the sphere from the values in the CAD file. Thus, the cross-sections will also vary slightly.

2. New questions: December 02, 2010

In your problem description you mention that "The hexagonal grids were located at 60° from the build direction." It seems to me that the hexagonal grids are actually 45° from the build direction. Can you clarify this? Are you testing at 60° from the build angle or along the axis of the hexagonal grid? They do not appear to be the same to me. Also, when can we expect to have the data from this loading at the third angle (the one that is not 90° or 0°). Thanks!
You are correct in finding that there is a 6 fold rotational axis located at 45 degrees between the build direction (z direction) and the perpendicular to the build direction (x-y plane). There are several of these occurrences that repeat based on the entire geometry of the sphere. If you observe the sphere perpendicular to the z direction (in the x-y plane) down the 6 fold axis, it contains the top and bottom of the build which is considered the build direction. The 60 degree rotation from the build direction is the testing direction. A graphical image depicting looking down the 6 fold rotation axis located in the x-y plane can be viewed in the gallery. You can expect the data generated along the 60 degree test direction to be released by 1630 EST on 02DEC10.

3. New questions: November 30, 2010

Have you considered splitting the prize money half-way for the sphere and cube? Most participants are likely to compile different models for both instances and as such the winner of the overall competition might not have the best model per instance. By splitting the prize money your objective would be met such that you reward the best models for each instance.
We thought a great deal about how to structure this competition, and came up with a single award for one winner for several reasons. Primarily, we wanted to incentivize every competitor to work on each 'half' of the challenge, regardless of which one they might have specialized talents that will enable them to solve one part or the other.

A valid submission for the 'final' configurations will require a description of _both_ models from each competitor, so we'll be able to see the model descriptions from everyone.

We'll definitely list the individual and team rankings for each half of the Challenge, but to take home the $50,000 prize, you have to get closest to both. That individual or team will get singular recognition for their model development.

If you're having difficulty with one part of the Challenge, we suggest you attempt to gather some collaborative support for your team. One way that might work is to ask for assistance on the DMACE Challenge Facebook page: www.facebook.com/DMACEchallenge. Or you can request to join another team already registered at www.DMACE.net, and explain what you'd be able to bring to the table as a team member.

4. New questions: November 29, 2010

Just FYI, in file 'Set 1- Current=1.7_Velocity=200_CrushAngle=0.xls', sheet B5, the max displacement value is not correct.
Thanks to one of our faithful competitors for pointing this out; the correct formula should be '=MAX(A:A)', resulting in a value of '9.8902 mm'. It doesn't affect the competition, but has been corrected and reposted on the Week 1 Data (Sphere) page.
Regarding the final set of spheres (hexagonal grid at 60 degree) for which participants will be predicting the maximum compressive load:
  • What would be the number of trials for this configuration?
  • Will the process parameters be a combination from the given current and velocity variables (i.e. 1 from the 9 possible combination), or would it be an inter-polation/extra-polation of the variables (Ex: 1.5 mA at 250 mm/sec beam velocity)
This is an excellent question! However, information on the final configurations of spheres and cubes will not be made available until Friday, 03DEC10.
What number am I trying to guess for the spheres?
You are trying to guess the value contained in cell C4 of the "Summary tab" for each set of spheres. The task is to guess the maximum compressive load for a given combination of digital manufacturing input settings, which is determined here by averaging the maximum compressive load of several sphere(s) tests made with that particular combination of digital manufacturing input settings.

While there does exist a compressive load in each set of spheres or cubes that is a 'maximum', the DMACE Challenge is to seek a correlation model that more closely matches the physics of the digital manufacturing process. That is, the 'average' maximum compressive load represents the range of values in testing _all_ of the resulting spheres and cubes, not just the single example of the one with the 'maximum'.

This has value in potentially reducing test and evaluation requirements. If an engineer can use the model to predict an 'average' maximum strength that has sufficient margin above requirements, you could rely on _all_ produced components to perform as desired without the need for extensive testing.
You say that “to be valid, your final entry must include your models and a brief description of the techniques used in the development of your models (PDF is preferred).” How brief of a description - is there such a thing as too brief (like a minimum number of words)?
The DMACE Challenge has not set a minimum for the model submission, but the intent is to capture enough understanding about how each competitor or team modeled the correlations to expand the basic science of the physics of digital manufacturing. Tell us how you came up with your estimations and why you used the math/models you developed. When you win, your submission will be known throughout the world as winning the DARPA DMACE Challenge, so show us how much work went in to winning the $50,000.
In the final challenge, you "won't be giving rankings for estimations" - will you be allowing multiple submissions and providing the SMD for each submission? I was able to submit an entry multiple times for Practice #2. Using a simple binary search for both values in isolation, one can find the exact values with 2 decimal points, if given an infinite number of submissions and the SMB for each submission.
Practice #3 will very closely match how the Final Challenge will work. It will require a sphere and a cube estimation AND an uploaded file. This file can be a bare-bones description of your model (PDF preferred) but is really testing our ability to receive the files that we’re requiring for the Final Challenge. Yes, since we posted results in real-time on the leader board for #1 and #2 (allowing some people to 'game the system') the same won't be true for #3 or the final. Contestants will enter their estimates and see _nothing_ about their SMD or how they rank in the standings until the close of Practice #3 or the Final Challenge, respectively. You can still enter as many submissions as you want; however, only the last estimates you enter will count, and each later submission will change the ‘time stamp’ you’ll get for your entry. For the Final Challenge, the closest SMD will win; in the case of a tie, the competitor with the earliest submission with that SMD will win.

5. New questions: November 18, 2010

As a team leader, how can I add competitors to my team?
You can either:

1) Log in and go to the “Manage My Team” option, entering them manually. (You will have to enter their e-mail and then re-enter it for confirmation).

2) Have them register on their own and then they can go to the “View Teams” option and click the link for your team. You will receive a request from the website, allowing you to accept them on your team or reject them from joining.

By the way, you can also add a team logo to display on the Team Page (inappropriate images will result in banning the team leader from the competition, so please use good etiquette!)

6. New questions: November 12, 2010

What materials and machines are you using for the digital manufacturing process in the DMACE Challenge?
For the cube structures, dog bones and compression samples, the data is generated with a FORTUS 400mc machine and ABS-M30 production-grade thermoplastic. For the sphere structures, an Arcam A2 machine and Ti6Al4V Titanium Alloy powder is being used for digital manufacturing and testing. Files with this information and more can be found in the data pages.
Are you going to publish additional information (material properties, geometry, etc..) pertaining to each run?
We will be posting the Computer-Aided Design (CAD) for the sphere and cube structures in the near future.

As far as the ‘cube’ material properties go, the data being released using the ABS-M30 thermoplastic starts with simple compression samples in order to give you information on the material properties after the samples are digitally manufactured. Next, we will provide similar test data from ASTM ‘dog bones’ to expand your understanding of how the thermoplastic performs. Then, we will provide a CAD file for the cube structure (which is based on an international-standard ‘cube satellite’ design) and compression test data on those structures.

The ‘sphere’ structure will use variations on the speed and the power settings of the electron beam that melts the titanium alloy powder level by level in the 3D CAD file design. The compression tests will vary the angle that the spheres are crushed – that is, a number will be compressed on the same axis that they were assembled, and another set will be compressed after rotating them 90 degrees from the structure ‘vertical’.

7. General

What is DARPA?
The Defense Advanced Research Projects Agency (DARPA) is the central research and development (R&D) organization for the United States Department of Defense (DoD). DARPA manages and directs selected basic and applied R&D projects for DoD and pursues research and technology where both risk and payoff are very high and where success may provide dramatic advances for traditional and military roles and missions.
What is digital manufacturing?
Digital manufacturing (DM) involves the use of computers and “printers” that create physical components from the direct application of raw materials in accordance with a computer-assisted design (CAD) layout. DM techniques have the potential for the development of components that were previously very difficult or expensive to produce using conventional techniques.
What is the DARPA DMACE Challenge?
Under the Digital Manufacturing Analysis, Correlation and Estimation (DMACE) (pronounced “DEE-mace”) Challenge, DARPA will digitally manufacture several complex structures and then conduct a series of structural load tests. Data from the manufacture and load tests will be posted on the Challenge website, http://www.DMACE.net. Participants will be challenged to develop a correlation model that accurately correlates DM machine inputs to output structural test data. Participant models will be evaluated by their ability to predict the test results of the final DM structures. The model that most accurately predicts the final test results wins the Challenge.
Why is DARPA investigating DM applications?
Manufacturing small quantities of complex system components can be costly and time consuming. DM holds the promise of lowering cost and time, both of which are critical concerns for the complex systems required to support DARPA mission objectives. The Challenge will be used to determine whether predictive correlations exist between DM settings and resultant product structural properties.
When will the Challenge take place?
Data for model development will be released incrementally beginning on or about October 29, 2010, and will continue through December 1, 2010. The final parameter change requiring correlation prediction will be posted via the Challenge website on December 3, 2010. Participants will have until 4:30 PM EST December 6, 2010, to submit their prediction for the final structures. During the week of December 6, 2010, the final samples will be tested and the winner announced.

8. Eligibility

Who is eligible to enter the competition?
The Challenge is open to individuals, and teams of individuals, of all ages. Participants under 18 years of age may be required to obtain the consent of a guardian and/or meet other applicable legal requirements as a prerequisite to accepting the Challenge prize.

DARPA employees and family members, as well as DARPA support contractors and their family members, are ineligible to participate in the Challenge. Similarly, Oak Ridge National Laboratory and U.S. Naval Postgraduate School employees, support contractors, and the dependents thereof who are involved with the design, manufacturing, testing and analysis of DMACE Challenge content or materials are ineligible to participate in the Challenge.

Non-U.S. citizens require a taxpayer identification number (TIN) to receive the prize. A TIN is obtainable upon request from the U.S. Government. Additional information is available on the U.S. Internal Revenue Service website at www.IRS.gov.
Do I have to register to participate?
Yes. Participants must register on the DMACE Challenge website (http://www.DMACE.net) before they submit entries. Registration will open on or about October 29, 2010. Registrants must have a working email address to complete the registration process.
Is collaboration authorized in developing my entry?
Collaboration is authorized and encouraged for developing your entry.
May a participant be on more than one team?
Participants may be on more than one team. However, an individual participant or team may not submit more than one entry.
Will DARPA provide funding to solve this Challenge?
No funding or other monetary assistance (other than the prize to the winner) will be offered.
Can an email address, website or other account include “DARPA” in its name? Can I use the DARPA logo on my website?
No. DARPA’s name and logo are protected under Federal law, and actions that could constitute trademark infringement or trademark dilution should be avoided. Participants should also avoid actions that might suggest or imply official endorsement by DARPA or the U.S. Government.
As a team leader, can others join my team after I have registered?
Yes. When a team registers, it must list a team leader and team name. Team leaders determine the individual participants on their team. The team leader will be the primary point of contact for the team.

9. How the DMACE Challenge Works

What are the structures that will be manufactured and tested?
DARPA will digitally manufacture structures composed of metallic and polymer materials. The metallic structures will be hollow mesh spheres. The polymer structures will consist of complex cubes and 2-dimensional test specimens (e.g., coupons). The 2-dimensional test specimens will be used to investigate basic material properties of the polymer. Further information on the materials used to manufacture the structure will be posted to the DMACE Challenge website (http://www.DMACE.net).
How will the structures be tested?
All DM structures will be tested in accordance with conventionally accepted methods. Three- dimensional structure testing may include, but is not limited to, compressive load testing. The simple 2-dimensional polymer test specimen testing may include, but is not limited to, compressive and tensile load testing. Data from the structural load testing will be provided via the Challenge website.
What type of models can I use?
The technique of model development is up to the participants. A model may include fundamentals from one or a combination of mathematical, stochastic, structural finite element or any of the many other available techniques. It is up to the ingenuity and creativity of the participants.
What information will be provided so I can develop my models?
Data will be provided for both DM process input settings and the corresponding output structural load testing. Input setting data may include, but is not limited to, device control parameters, material composition, and CAD files. Input setting values will be changed over the course of data collection to provide a robust cross-section of data for model development. Output data may include, but is not limited to structural load test results such as stiffness, strength and displacement data. Data may be provided in a variety of formats, tabular and graphical.

All data sets will be provided on the Challenge website to registered participants.
In what format will data be provided?
Data will be provided in International System (SI) units in the form of tables and charts.
How will the models be evaluated?
For the final evaluation of participant models, several DM input settings will be changed. These may include one or a combination of, and are not limited to, the device control parameters and the complexity of the structure. For the final test configurations, the DM input settings will be altered to previously undemonstrated conditions. Entrants will be required to develop models for two different structures.
In what format do participants need to submit their entries?
Participant entries shall be submitted in SI units. Further details regarding the level of precision will be posted on the Challenge website.
What is the deadline to submit predictions?
Entries may be submitted until 4:30 PM EST December 6, 2010, or later if DARPA extends the submission deadline.
Where do I submit my final predictions for the DMACE Challenge?
Final predictions for the Challenge should be submitted on the DMACE Challenge website.
What makes a valid entry for the Challenge?
To be valid, an entry must include the predicted result of the structural test for each component; i.e., there will be a prediction for the metallic material structure and the polymer material structure. In addition, a valid entry must include the models and a brief description of the techniques used in the development of the participant’s models.
How will the final winner be decided for the Challenge?
The participant with the lowest sample Mahalanobis distance will be declared the winner. The formulation for determining the sample Mahalanobis distance will be disseminated on the DMACE Challenge website.

In the event of two entries with the same result, the entry with the earliest submission time will be declared the winner.
Who should I contact if I have a question regarding the website, the DMACE Challenge, or the rules?
Questions should be sent to the DMACE Challenge webmaster at webmaster@DMACE.net.
What social media sites will DARPA use to provide information about the Challenge?
In addition to the DMACE Challenge website (http://www.DMACE.net), DARPA will use Twitter and Facebook to provide information about the Challenge.
How can I request additional information to develop my model?
Requests may be sent to the DMACE Challenge webmaster at webmaster@DMACE.net. Requests for additional information may also be submitted via Twitter and Facebook where other participants can actively participate in discussion on the subject in question.
I am having difficulty developing a model to predict one of the two results. Can I submit a prediction for only one of the two test results?
Entrants may submit an entry for only one of the two structural tests; however the winning entry must have a prediction for both structural tests.
How many times can I submit an entry to the DMACE challenge?
Only one submission per team or individual participant is allowed.
Are automated entries allowed?
Automated entries are not allowed and will disqualify that participant. Automated entries include all automatic, programmed, robotic, or similar means of entry that are not based on a participant-developed correlation model.
Are emailed entries authorized?
Emailed entries are authorized; however, the Challenge website is the preferred method to submit entries.
How will the winner be announced?
The winner will be announced via email, an official DARPA news release, and updates on Facebook and Twitter.
Will the final data results be released so I can see how my model compared to others?
Final results will be released on the DMACE Challenge website (http://www.DMACE.net). Information released to the public will be limited to the individual or team names and final predictions for all entrants.

10. Prizes

What if multiple entrants submit the same final answer?
In the event of a tie, the valid entry with the earliest submission will be the winner.
Can an individual represent a large organization and distribute the prize money to a team?
Can the prize money be paid to my limited liability corporation rather than to me personally?