Subscribe Bookmark
arati_mejdal

Staff

Joined:

May 21, 2014

Catapulting to Success with Design of Experiments

Sarah Flexman is an 11-year-old middle-school student who uses JMP … for design of experiments … with proven success.


A student at East Garner Magnet Middle School in North Carolina, she is an accomplished singer, flute player and athlete. Her father, Greg Flexman, uses JMP in his work leading the Processes and Risk Analysis group at Talecris Biotherapeutics.


At the end of 2010, Sarah had decided to take part in Storm the Castle, one of the events offered in the statewide science Olympiad competition. This particular challenge was to design, build and launch a model trebuchet, which is a medieval-style catapult for hurling heavy stones (see below).


Sarah Flexman with her model trebuchet


With only six weeks to construct her entry into the NC Science Olympiad, Greg taught Sarah how to use JMP for design of experiments (DOE). They both stopped by my office recently to tell me about Sarah’s project.


An open event in the Science Olympiad, the trebuchet competition includes many requirements. For instance, the trebuchet had to fit in 75 cm cube. Weight ranges for the counterweight and projectile were given, but the specific values would not be known until the day of the competition, when the competitors would have to predict the outcome and then be judged on both distance and accuracy to their prediction.


Sarah found a trebuchet design she liked online and calculated the ratios for the dimensions on paper to fit. To build it, she used mostly materials her family had at home already. For instance, the wheels came from a child’s scooter and the throwing arm was part of an old tent pole. Sarah did all of the construction on her own, including cutting the tent pole and pieces of plastic. Other parts she put together included an axle, sling, a track for the sling, side rails and the fulcrum (pivot point).


So how much help did she get from her dad?


“He helped me with the DOE,” Sarah said.


“I got her started with DOE. She used the Custom Designer and filled in the blanks. All I told her was that it should be a third-order model with interactions. The Custom Designer generates a data table with the X parameter values and ‘fit model’ script all ready to go,” Greg said.


Using DOE, Sarah optimized to maximize distance using the model. This predicted how the trebuchet would do and how the various factors affected the outcome.


Here’s Sarah’s whole process: She built the trebuchet, tested it, used JMP for DOE during optimization, changed the hook angle and sling to improve performance, did more tests, entered this new data, reran the model, and made her final prediction graphs. The variables in her DOE were string length, counterweight and projectile weight, and she optimized for distance – that is, how far the projectile would go.


JMP Prediction Profiler


“Rather than doing 125 tests because we have three variables with five levels each, DOE found a way for us to perform only 26 tests and get approximately the same results. I typed in the results, ran the model and used the JMP Profiler. I understood how the variables predicted the outcome and found several patterns,” she explained.


So, JMP made a big difference.


“Because I started several months late, we were tight on time. JMP saved us a lot of time drawing the graphs. We would have had to draw those by hand. The DOE was also huge in characterizing the trebuchet. The whole thing would have taken much longer, and I had only six weeks,” she said.


Sarah said she didn’t find JMP particularly hard to use. “Generally, it was very easy once I started using it. You just plug in the information,” she said.


On the day of the competition, Sarah brought her trebuchet to an arena at Campbell University in Buies Creek. The competitors were given the specifics of projectile weight (40 grams) and the counterweight (2.5 kilograms). Given these values, Sarah predicted that her trebuchet would shoot 16 meters. Like the other competitors, she had five minutes in which to take three shots. The competition space only offered a maximum of 13-meter target placement.


 Sarah Flexman launches trebuchet at NC Science Olympiad competition


Her trebuchet was too powerful. In all three attempts, she hit the same seat in the arena stands, well beyond the 13-meter target. So unfortunately, her accuracy score was deducted because the projectile landed several meters beyond the target, as you can see in the photo below.


Trebuchet launches projectile into arena stands


“I should have rehearsed the five-minute run,” she noted.


Nevertheless, Sarah took 2nd place in the regional competition for middle school students, an impressive result. Bravo, Sarah!


She said building the trebuchet and competing in the Science Olympiad was a fun experience overall.


“I hadn’t done any building like that. The whole day was fun. It was a very open learning environment. You were experimenting with things you had never done before. I would definitely do it again,” Sarah said.


And she will – next year.


(Note: All photos provided by Greg Flexman.)

7 Comments
Community Member

Liza Lucas wrote:

Nice job, Sarah! This is interesting work, and I'd say the "apple doesn't fall far from the tree!"

Community Member

Trish Mullins wrote:

I was a judge for the Experimental Design event at the Buies Creek Science Olympiad Regional Tournament held at Campbell University on February 19. It is wonderful to see all the cool things that kids think up each year. Go Sarah!

Storm the Castle is not exactly an "open event" as stated in the article. Your school must enter a team. Schools or home school groups can form teams of up to 18 students to compete in pairs or threes across 23 different events. It takes a lot of adult/parent help for kids to learn what is needed for these events. Go Greg, for taking this kind of time to help!

Trish Mullins

Science Olympiad Coordinator

Cardinal GIbbons High School

Community Member

Steve wrote:

Outstanding work Sarah!

I have a first grade daughter and we have just started using math software on our home computer for school projects. I look forward tograduating with her to statistical ones

I do many optimization projects for my company and we have often found your experience to be true. The design solution is optimized well and has excellent predictability within th erange of tested parameters. In the real world application however, things/effects you don't expect can reach out to "bite" you. In your case, the time required for the shots and the 13m target distance limit.

Great work and I hope you do even better next year!

Steve

Community Member

Sarah Flexman wrote:

Yes I could have shortened the strings. I hadn't rehearsed the five minute run, though I should have. Therefore I was nervous about time and being disqualified, but yes that would have shortened the shot to around 13 meters.

Community Member

John HUnter wrote:

Thanks for the post, it is great to read about kids using DoE. I remembering being taught about it at that age myself by my father - http://williamghunter.net/ Keep up the good work, Sarah.

Community Member

Louis Valente wrote:

Wow Sarah this is excellent! I spent an entire career at Eastman Kodak getting scientists to embrace the Design of Experiment approach to investigations and you have already embarked on that journey. I am very impressed and you give me hope that future scientists will embrace such a powerful methodology at an earlier stage in their careers. Keep up the excellent work because the stars are within your reach!

Community Member

Daniel wrote:

Would cutting the sting length have helped? She may not have been allowed to modify the trebuchet during the competition, but her tests certainly could be used predict the effect of shortening her string to cut the distance down to 13 meters.

In any case, congratulations are in order! Great application of DOE.