Discover a new florescent material from patent data
	Design a new superconductor from experiment data
	Improve the performance of a VPTC ceramic semiconductor
	Optimizing the Production of Synthetic Rubber
	Carbon fiber reinforced, resin-based composite materials
	Bi2O3-containing High Tc superconductors
	Electrode materials of Ni/H batteries
	High temperature, SiC-based structural ceramics
	High-polymers - PVC, synthetic fiber, polyethylene, synthetic rubber, …
	High energy materials
	Semi-conductor devices
	MOCVD method of III-V compound film
	Polymers hydro-cracking vapor recovery platinum reforming delayed cooking de-waxing vinyl acetate polypropylene jet fuel (modified the recipe, yield 87% -> 94%, +6,000 tons/yr) increase life of catalyst in polyvinyl plants (catalyst cost $1.2MM) VC, polyformaldhyde

Intelligent Materials Design by Data Mining

Summary:
(abstract of a keynote presentation by Zaptron at IPMM'99 - the 2nd International Conference on Intelligent Processing and Manufacturing of Materials, Honolulu, Hawaii, July 11-15, 1999)

Though human beings have a long history of developing, manufacturing and using materials, drugs and chemical products, we continue to look into new materials and substances to meet the ever-increasing needs in industry, agriculture, health, and national defense.

Historically, the exploration and development of new materials (substances) are based on experimental methods, or the trial-and-error methods. he 1950s trend in materials and molecolar design was focused on studies of relationships between material structure and material property by quantum chemistry. Later, people adopted more sophisticated methods, such as artifical intelligence, data
bases and computer imformation processing, to build expert systems. The computer-aided methods have been the main stream in the field of advanced materials design. Some of these expert systems are also used for the optimal control in the manufacturing of materials. In addition, as the newewst generation in materials research and process control, the so called intelligent processing of materials
is now beeing adopted in the manufacturing of various new materials.

Today, materials design is not only a hot research topic at institutes, but also a critical focus in industry. A completely new methodology is heing developed in the research and development of
advanced materials and substances. There are three levels of computerized materials design:

Level 1. use quantum chemistry, solid physics, structural chemistry to study relationships between material microstructure and property

Level 2. explore the development of new alloys, ceramics and semiconductor materials by using phase diagram, thermodynamics
and kinetics.

Level 3. apply pattern recognition, data mining, neural nets, and genetic algorithms to optimize the manufacturing, processing and property of materials. Very often this is accomplished with the aid of data base, knowledge base & knowledge discovery.

Our work focuses on Level 3 in intelligent materials design - application of hyperspace data mining to the optimal design of materials. We will introduce an expert system, MasterMiner™, a software suite that is comprised of knowledge base, data base, data mining algorithms, artifical neural network, and generic algorithm.

The data mining module is used to separate data into subspaces so that a feasible mathematical model in the original parameter space can be obtained. An important step in data mining is the separability test, whereby a good separability leads to a feasible model in a subspace and vice versa. Our experience indicates that about 20% of the field data are not separable, and these data show no pattern for recognition.

Once a data pattern has been recognized, an artifical neural network can be trained to find the mathematical model for the pattern. Then a genetic algorithm can be developed to obtain the best solution by searching for the global optimum. Finally, the knowledge obtained by the expert system, including the mapping graphics by data mining, criterion equations, trained neural nets and the optimum, is stored in a materials data/knowledge base for later use. A number of real-world design examples will be given to show the efficacy of the proposed method.

LIST OF TERMS USED

- data, database, data mining, data patterns
- steps in intelligent materials design
- pattern classification by data mining (separablity test)
- feature selection and reduction
- hyperspace modelling using polyhydrons, projection, neural nets and GA
- optimal design and extrapolation to new materials
- examples: superconductor, VPTC ceramic semiconductor, high brightness phosphor, etc.

pattern recognition& data mining Innovations in Materials Design materials & process discovery new materials/process knowledge Foreword of International Conference on Intelligent Process in Manufacturing of Materials (IPMM'99, Hawaii) Materials, and therein materials design, has evolved over two millennia from the simple to complex macro-design of mixtures - monolithic to composite materials. The trend today is thin films for various applications, and the research emphasis is toward more computationally tractable  'atomic-scale' design of lattices, surfaces and interfaces.  In general, the pace of all materials research is being challenged, particularly from an affordability perspective.  As a consequence, a world-wide pursuit of more efficient and accurate prediction methods of ‘yet-to-be-made’ materials is becoming a preeminent materials research frontier.  One, currently very popular, approach to materials design is to utilize existing materials data to predict properties of yet-to-be-made materials.  Because of the vast amounts and varying quality of this information, the use of search-based methods for augmenting more analytic approaches is receiving increasing attention.

Toward the Future: Innovations in Materials Design
Shuichi Iwata
University of Tokyo

With the advent of personal computing and the internet, the opportunities for realizing new ideas and innovations are virtually limitless.  In the context of materials development, an area which is pervasive and impacts nearly all aspects of our lives, these opportunities are potentially very profound.  Materials design, as a profession and a field of research, has had an enduring legacy of contributions, involving methods ranging from first principle calculations to mesoscopic and macroscopic techniques, and more recently empirical or data-driven approaches.

Innovation is often associated with a 'leap' in performance or breakthrough from one equilibrium point to another new point with new values.  Innovations require 'something' new as well as incremental efforts based upon a feasible and rational plan.   It is a dynamic process, driven by a 'sixth' sense and a human desire to discover.   This human aspiration to innovate will be driven by and will leverage an ‘exploding’ information technology age which will foster evermore creative environments for design.

Users of these environments will use tools which evolve and adapt to the user community.  The speed and ease of use of such environments will accelerate the fusion and synergy of varying opinions and ideas, while minimizing expensive trial and error.   New insights, however unorthodox, may be explored more rapidly, and at minimal expense, via virtual design environments which enable the simulation and modeling of artifacts regardless of their complexity. In this context, we have organized this ‘Innovations in Materials Design’ session to stimulate your awareness of future directions for materials design, and more generally, we want you to reconsider the 'richness' of the computing resources, data, knowledge, methods, and communication technology available to you.