Difference between revisions of "The leaky tank mystery 2024"

From Derek
Jump to: navigation, search
(→‎Honours students)
(→‎General project description)
Line 12: Line 12:
  
 
==General project description==
 
==General project description==
−
In this project is about human identification. In times of war and natural disaster there is a high demand for the ability to identify human remains. Also in the area of policing crime, the need to forensically identify remains is critical.  Also human identification can extend to humans that are alive: for example reuniting an adopted child with a birth family. Children that do not know their origins for a number of reasons (eg. human trafficking of babies) can also benefit. 
+
Imagine one of those railcars that has a big horizontal water tank.  Pretend it is frictionless. Imagine the tank is full.  Imaging there's a hole at the top to let air in. Imagine the tank is leaking through a small hole on the left bottom side of the tank.  Assume the jet of water is vertically downward. Does the resulting change in momentum push the railcar to the left or the right?
  
−
The idea of this project is to advance the techniques of human identification by focusing on a difficult case that that is too hard for current techniques. You can read the details about the dead body and the circumstances [http://en.wikipedia.org/wiki/Taman_Shud_Case here.]
+
This is a very famous problem that has never been properly solved. We need finally put it to rest.  We will give you papers showing attempted solutions that contradict each other. You may try to search for more papers on the topic.
  
−
We also want you to bring the skills of an electrical engineer to bear on the area of e-forensics and bioinformatics see if you can apply these to an important area of the case, ie. analyzing degraded DNA sequences.
+
The experiment is almost impossible to do in practice as the forces are smaller than the friction you get in real life. This explains why there are contradictory solutions.  This illustrates an important principle in engineering that your can never be sure of the theory until you do the experiment.
 +
 
 +
But the modern world of engineering now has simulation tools that we trust and are as good as an experiment provided we don't feed it with wild parameters outside of the range that models are valid.
 +
 
 +
We want you to do the "experiment" in simulation using Comsol and/or Ansys.  The Faculty has licences. Find the answer. Then go and find which theory paper got it correct.  Then nicely explain that theory yourself, written out clearly in your own way.
 +
 
 +
This is a beautiful classic problem and it will be a lot of fun. If you can think of a clever way to actually do a low friction experiment and video it, this will be a cherry on that cake and your name will go down in history. Suggest you design a rectangular tank and suspend it hanging with four cords attached to the corners.  You'll need to design a release valve for the water that is actuated by a remote (because you can't touch the tank). Suggest you bounce a laser pointer off the tank as a way of measuring deflection.  However, you need to fully simulate the tank first to get all the size parameters right to optimize the size
 +
of the deflection.
  
 
==Specific tasks==
 
==Specific tasks==

Revision as of 18:02, 19 May 2024

Supervisors

Honours students

Project guidelines

General project description

Imagine one of those railcars that has a big horizontal water tank. Pretend it is frictionless. Imagine the tank is full. Imaging there's a hole at the top to let air in. Imagine the tank is leaking through a small hole on the left bottom side of the tank. Assume the jet of water is vertically downward. Does the resulting change in momentum push the railcar to the left or the right?

This is a very famous problem that has never been properly solved. We need finally put it to rest. We will give you papers showing attempted solutions that contradict each other. You may try to search for more papers on the topic.

The experiment is almost impossible to do in practice as the forces are smaller than the friction you get in real life. This explains why there are contradictory solutions. This illustrates an important principle in engineering that your can never be sure of the theory until you do the experiment.

But the modern world of engineering now has simulation tools that we trust and are as good as an experiment provided we don't feed it with wild parameters outside of the range that models are valid.

We want you to do the "experiment" in simulation using Comsol and/or Ansys. The Faculty has licences. Find the answer. Then go and find which theory paper got it correct. Then nicely explain that theory yourself, written out clearly in your own way.

This is a beautiful classic problem and it will be a lot of fun. If you can think of a clever way to actually do a low friction experiment and video it, this will be a cherry on that cake and your name will go down in history. Suggest you design a rectangular tank and suspend it hanging with four cords attached to the corners. You'll need to design a release valve for the water that is actuated by a remote (because you can't touch the tank). Suggest you bounce a laser pointer off the tank as a way of measuring deflection. However, you need to fully simulate the tank first to get all the size parameters right to optimize the size of the deflection.

Specific tasks

We have supplied you with a file of about 700,000 rs and i numbers and the corresponding SNPs for the Somerton Man. Unfortunately there's a very high drop out rate. The idea is to groundtruth this data, characterize it, and squeeze out of it any information you can. The idea is to find out what the data can tell us and also what it definitely fails at telling us. We need to know both. Here are some ideas to get you started:

  • Write a script to count the rs and i numbers in each chromosome. Count the base pairs in each chromosome. Tabulate these results. Do this quickly.
  • Create a synthetic "random human" file to get some idea what random SNPs do. Upload this in research mode on Gedmatch Genesis and test its characteristics. Does it actually link to any humans?
  • Create a synthetic human that say has all A's or a short sequence of SNPs that is repeated over and over periodically. How do these files behave?
  • Upload the Somerton Man in research mode. Find out which chromosomes are under the minimum required based papers. Try many different ways of padding these base pairs with dummy data in order to meet the minimum. You are looking to find the padding method that influences the results the least.
  • Using a known good DNA reference file, deliberately drop out those SNPs that are missing in the Somerton Man. As you know the groundtruth in this case, you can investigate which padding methods failed and which produced partial good results.
  • Another padding trick is you might introduce homozygous pairs (AA, GG, TT, CC) in various places to beef up the number of SNPs. These segments might distort the relationship estimates, but if you keep them to smaller segments, they might not make a difference. You can test this out on the good DNA file that has been deliberately stripped.
  • Further to padding by just the minimum amount, you can then investigate if you can pad by more by finding SNPs that are common to males or have a high liklihood in males. You can test this on good DNA files.
  • You can go to a SNP database that is indexed by rs and i numbers and find the affect each SNP has (eg. if it affects eyecolour, a disease, or whatever). You can then create you own file containing all the descriptions for the Somerton Man. Then you can write your own software to datamine this file for interesting physiological features, characteristics, and diseases.

Deliverables

Semester 1

Semester 2

Weekly progress and questions

This is where you record your progress and ask questions. Make sure you update this every week.

Approach and methodology

We expect you to take a structured approach to both the validation of last year's results, and the writing of the software. You should carefully design the big-picture high-level view of the software modules, and the relationships and interfaces between them. Think also about the data transformations needed.

Possible extension

  • If there is time we can get hold of the original FASTQ file and perform statistical tests on it. And you can attempt statistical imputation to infer missing SNPs. Idea behind imputation is like what we do in electronic engineering when we do error correction in digital communication systems. For example if a communication system transmits the sentence "the cat sat on the mat" and we receive "th* c*t s*t *n th* m*t" with drop outs in information, we can still reconstruct the sentence. You can think of a DNA file as being like a long message that also suffers drop outs, and so you can use statistical techniques to infer values. We electrical engineering we call this "error correction" but in bioinformatics this is called "imputation."

Expectations

We don't really expect you to find the killer or identify the Somerton Man, though that would be cool if you do and you'll become very famous overnight. To get good marks we expect you to show a logical engineering approach to squeezing information out of the data, and using known good files to always groundtruth each idea.

It is perfectly acceptable to have a long list of ideas that didn't work, provide they are carefully tested in a structured way. Finding things that don't work is part of the scientific process. When we don't know what is supposed to work or not, being able to eliminate ideas that don't work is still very exciting and worthwhile.

  • It is important to regularly see your main supervisors. Don't let more than 2 week go by without them seeing your face briefly.
  • You should be making at least one formal progress meeting with supervisors per month. It does not strictly have to be exactly a month, but roughly each month you should be in a position to show some progress and have some problems and difficulties to discuss. On the other hand the meetings can be very frequent in periods when you have a lot of activity and progress to show.
  • The onus is on you to drive the meetings, make the appointments, and set them up.

Relationship to possible career path

Whilst the project is fascinating as you'll learn about a specific murder case—and we do want you to have a lot of fun with it—the project does have a hard-core serious engineering side. It will familiarize you with techniques in information theory, probability, statistics, programming, bioinformatics, and datamining. It will also improve your software skills. So go ahead and have fun with this, but keep your eye on the bigger engineering picture and try to build up an appreciation of why these techniques are useful to our industry. Now go find that killer...this message will self-destruct in five seconds :-)

See also

References and useful resources

If you find any useful external links and resources, list them here:

Back

Retrieved from "https://www1.eleceng.adelaide.edu.au/personal/dabbott/wiki/index.php?title=The_leaky_tank_mystery_2024&oldid=23423"