Elmer: Open Source Finite Element Software

If you're into science or design engineering, you should probably check out Elmer. Among a host of other great applications, this free, Finland-born software can help you design airplanes, predict the temperature distribution of a heat exchanger, and do your quantum mechanics homework.

At its core, Elmer is an open source finite element solver of partial differential equations. Development of Elmer began in 1995 as a collaboration between Finnish universities, research institutes and private industry, and was primarily developed by Finland's CSC IT Center for Science. Elmer was released as open source in 2005. According to the Elmer FAQ page, Elmer has hundreds of regular users worldwide and thousands of Elmer test users annually.

Click here for a 3-minute video introduction to Elmer.

Elmer processes partial differential equations in a descrete form, and handles coupled systems, non-linearities, and time-dependencies. The Elmer GUI allows the user to either import meshes or create simple ones in a variety of file types, and generates output in .grd, .mesh, and .ep files. The source code of Elmer is written in Fortran 90, C, and C++, and is distributed under the GNU Public License (GPL). The Elmer source code is here on sourceforge.

Here is a 10-minute Elmer tutorial from Elmer's YouTube Channel.


Thank you, Finland. It appears that your song needs a new verse.

Can we Make Science Cheaper by Crowd-Sourcing Experiments?

I admit that crowd sourcing science experiments sounds a little risky.  Since it's difficult to be fired from a job you do for free, contributors to crowd-sourced science experiments would have less incentive to do precise work or to keep their paradigms and biases from influencing the results they observe and report. Crowd sourcing science could certainly introduce unexpected and undocumented variables. On the other hand, large data sets are valuable.  And there may be a side benefit to gathering large data sets from an un-characterized group of real people living real lives in real homes - the data may lead to conclusions that are more directly applicable to the populations we are trying to learn about.

We can test the viability of crowd-sourced data collection by comparing conclusions drawn by crowd-sourced experimentation to the conclusions drawn from traditionally collected data. One example of crowd-sourced data collection is fuelly.com, where people report actual gas mileage for their vehicles. I did a quick spot check for the V6 Toyota Camry sedan. It would be an interesting exercise to repeat this check for a large group of vehicles.

On fuelly.com, the 119 participating drivers of V6 Toyota Camry sedans submitted fuel economy data for thousands of fill-ups. For this vehicle make and model, the most frequently reported gas mileage is 24 mpg, with a roughly bell curve shaped distribution of reported mileage ranging from a low of 16 mpg to a high of 32 mpg. According to fueleconomy.gov, fuel economy is measured for pre-production prototypes of new cars by the manufacturer using standardized test procedures specified by federal law. The EPA reviews the test results and spot checks 10 - 15% of them. This methodology predicted that the 2011 V6 Toyota Camry sedan would yield 20 mpg city, 29 mpg freeway, and 23 mpg combined.

In this case, the results are pretty close, but Camry drivers on fuelly.com got slightly better mileage than expected based on EPA regulated test data.  I saw lots of tips on fuelly.com for getting better gas mileage...I suppose it's possible that users of fuelly.com more frequently exhibit a bias toward maximizing fuel economy than the EPA testing procedures account for.