two math apps

Upload: jaymaj219249

Post on 03-Jun-2018

221 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 Two Math Apps

    1/14

    Two Math Apps

    Jayanta Majumder

  • 8/12/2019 Two Math Apps

    2/14

    Geometric CAD

    Engineering calculations- Stress distribution

    - Fluid flow

    - Cast solidification

    - Layout optimisation

    - Digital mock-up

    - Heat transfer

    Free Geometric CAD tools:

    http://www.123dapp.com/

    (Dont end up being a user)

    http://www.123dapp.com/http://www.123dapp.com/
  • 8/12/2019 Two Math Apps

    3/14

    Representation of 3d objectsB-Rep = Geometry + Topology

    Calculation engines :

    - Function minimisation

    - Root finding

    - Linear algebra

  • 8/12/2019 Two Math Apps

    4/14

    Shapes come from math

    functions

    (x,y,z) = ( 4+0.3*sin(7*u)+cos(5*v))*cos(u)*cos(v),

    4+0.3*sin(7*u)+cos(5*v))*cos(u)*sin(v),

    2*(4+0.3*sin(7*u)+cos(5*v))*sin(u) )

    (x,y,z) = ((3+0.2*sin(6*(u+v)+1))*cos(u)*cos(v),

    (3+0.2*sin(6*(u+v)+1))*cos(u)*sin(v),

    (3+0.2*sin(6*(u+v)+1))*sin(u))

    In practice the functions are

    mostly polynomials in

    entertainment and industrial

    product design and proceduralfor machine design

  • 8/12/2019 Two Math Apps

    5/14

    3D printing from equations

    using Mathematicahttp://www.ms.unimelb.edu.au/~segerman/papers/3d_printed_visualisation.pdf

  • 8/12/2019 Two Math Apps

    6/14

    Realism of Rendering : State of the

    artSkillset :

    Numerical robustness CS/Algorithms in general

    Numerical stability

    Understanding limitations

    of floating point

    computation

    GEOMETRY

  • 8/12/2019 Two Math Apps

    7/14

    Empirical Models :

    Models based on data

    Empirical models i.e. the models that help build and maintain empires

    e.g. models of cannon-ball trajectory.

    Modern empires like Apple, Microsoft, Google, Exxon, Shell, Dow runtheir operations based on empirical models of their plants and processes.

    [Citation needed]

    in fact the said etymology is false.

  • 8/12/2019 Two Math Apps

    8/14

    Empirical vs. Mechanistic Models

    It is more satisfying to find mechanistic solution to a problem because itgives more reach and is less parochial.

    Todays empirical model can instigate tomorrows mechanistic model

    e.g. Balmer-series model for hydrogen-emission lines were explained ~30

    years later by Niels Bohrs model.

  • 8/12/2019 Two Math Apps

    9/14

    But that doesnt always happen

    Titius-Bode planetary distance model (1772)

    D(n) = ( 4 3 2)/10 AU

    where n =, 0, 1, 2, 3, 4, ..

    It was fairly accurate for

    planets observable in 1772, but

    doesnt work for Neptune and

    Pluto.

    No mechanistic significance

    has been found.

    l net

    Actual distance

    from the Sun

    (in AU)

    Distance by

    Titus-Bode

    Formula

    Mercury0.387 0.4

    Venus0.723 0.7

    Earth1 1

    Mars1.52 1.6

    Asteroid Belt!3.0 2.8

    Jupiter5.2 5.2

    Saturn9.55 10

    Uranus 19.2 19.6

  • 8/12/2019 Two Math Apps

    10/14

    Method of building empirical models

    Connecting dots, but in higher dimensions.

    Hypothesize a function of the input variables in terms of a few freeparameters.

    Tune the free parameters for which the aggregate of errors is as low as

    possible. The tuning is a maxima/minima problem.

    How to come up with model functions?

    - Some mechanistic reasoning

    - Random trials

    - Understanding of shapes of terms

    - Study of the residue distribution and residue contributors

    - Sum of sigmoids and radial functions (neural networks)

  • 8/12/2019 Two Math Apps

    11/14

    Linear Model

    In this special case, the tuning turns out to be the problem of solving a systemof linear equations.

    In general, we minimise the following function over the possible values of thefree parameters vector a :

    () =

    (; )

    Linear model has the special structure that :

    ;

    ()()

    Here , , , are terms involving the ith input vector.

    Please note that the model is linear only in terms of its free parameters, but theterms can be arbitrarily non-linear.

  • 8/12/2019 Two Math Apps

    12/14

    Model Accuracy Criteria

  • 8/12/2019 Two Math Apps

    13/14

    A 1-variable model with level

    snapped data

    100 ( )Where t is the immersed time as measured in days.

    Here () 1 log( +) in which S is chosen to be 100.

  • 8/12/2019 Two Math Apps

    14/14

    THANKS!

    The End