Pirika logo
JAVA, HTML5 & Chemistry Site

Top page of Pirika

Chemistry@Pirika
 Properties Estimation
 Polymer Science
 Chemical Engineering
 Molecular Orbital
 Chemo-Informatics
 Other Chemistry
 Academia
 DIY:Do It Yourself
 Programing

Hansen Solubility Parameter (HSP)
  Basic HSP
  Applications
  Polymer
  Bio, Medical, Cosmetic
  Environment
  Properties Estimation
  Analytical Chemistry
  Formulating for Cosmetics
  Other
  DIY:Do It Yourself

Other Writing

Ad Space for you

 

Ad Space for you

 

 

 

 

Last Update
22-Jul-2014

Chemistry@Pirika: Chemo-Informatics

Japanese page

I developed a lot of Chemo-Informatics tools such as Neural Network systems, Genetic Algorithm programs and SOM programs. I will introduce my tools here. Actually there are so many systems that developed by mathematician or statistician. But such system is not adequate for chemistry. Pirika aim to develop "Of the Chemist, By the Chemist, For the Chemist" Chemo-Informatics tools.

How Neural Network learn Boiling Point of Compounds?

The problems in NN learning

DataBase Uncertainty destroy Neural Network

DataBase Problems

Pattern recognition by neural network

SOM(self organization map)

 

NN Examples:

Design of Azo-benzene: 99.3.24

For example, if you want to design red color Azo-Benzene dye, how can you do that?

There are 10 possible substitute positions and you can substitute NH2,OH,NO2,CN,SMe so on(maybe more than 10). Calculate 1010 possible structure with MO method? And pick out only red Azo-Benzene? it take 1000 years !

Design of Glass: 99.3.24

In case of Design of Glass, there are so many possibilities of each component ratio. If you try to melt 10 kind of oxide in 0.1% notch, it exist (103 )9 composition.

Only you have to do is input mol% of oxides, then you can get properties in a second. Glass transition Temperature, Young Modulus, density Refractive Index, Abbe's Number and Coefficient of Thermal Expansion.

If you can calculate glass properties in a second, it takes 1027 second calculate all 10 kind of oxide in 0.1% notch composition. 1 year exist only 3.1536*107 second, so you need 1020 years to evaluate all. If the target glass properties are already determined, please try my Java applet. you are designing High young modulus glass, then input desire Young Modulus, Coefficient of Thermal Expansion, density and Glass transition Temperature then push start button. you will get new glass composition. If you want to develop patent free new composition of lead glass, which is 10% lighter than your competitor but same Tg, TEC, YM, then please try my applet.
For this problem ChemoInformatics works very well.