Strategic Goals
We aim to motivate advanced technologies in the field of data science, and foster scientific methodologies for innovative design and manufacturing. Various fields in design and manufacturing play a major role in our country’s fundamental industry, which is now facing a revolutionary period. Population reduction and globalization challenge the current industrial structure in Japan, resulting in a rapid lost of our global predominance in the manufacturing industry. Furthermore, countries all around the world have actively launched new data science programs for the next generation of design and manufacturing, such as, the Materials Genome Initiative and the Industry 4.0. Following the global trend is no longer an effective way to survive in the intensive power game around the world. We have to develop innovative thinking, for which data science can provide the essential tools. The Institute of Statistical Mathematics has decided to establish a data science research center for the purpose of creative design and manufacturing. We have accumulated state-of-the-art data science knowledge here, for instance, machine learning, optimization theory, data assimilation, Bayesian inference, materials informatics, etc. We are devoted to foster innovative methods for design and manufacturing based on co-created values through industry-academia collaboration. This is the mission of this center and our keywords are “be smart” and “creative design and manufacturing”.
Smart Manufacturing in the Perspective of Data Science
Current approach of material design relies on a classical cycle: initial guess based on researcher’s intuition and experience, property estimation using large scale simulation and experiment, revision of the design based on the results. Such a process requires a large amount of time and financial support to discover and produce a new material. On the other hand, recently, there are new attempts to substitute the heavy simulation and experiment by statistical models based on many existing data. This approach will achieve extremely efficient estimation of the material properties. The enormous cost and time required in a property test has limited the new material discovery within a small set of candidates. High throughput screening using data science based simulation techniques may significantly increase the probability of new breakthroughs in materials science. Application of such advanced methods to the classical design and manufacturing process will drastically reduce the R&D cost and time. This is the basic concept of smart manufacturing in the perspective of data science.
Creative Design and Manufacturing Process based on Data Science
We recognize the importance of being at the absolute leading edge position in the manufacturing industry. This cannot be done by data science alone. Most of the classical data science analysis tools are designed for interpolating predictions. Data science used to be a science of predictions based on pattern recognition from existing data. For example, we often assume that materials with similar chemical structure exhibit similar physical properties. However, by definition, new material is not likely to be similar to any of the existing materials. Combination of experiment, theory and data science methods is an essential step to a breakthrough in the current state. We adopt a stepwise approach to expand the region of accurate prediction of a statistical model. We achieve the goal by careful design of an optimal experiment or simulation schedule for new data points that efficiently improve an existing predictive model. We have accomplished preliminary success in materials science using an extrapolating prediction method based on a data science approach, and fostered new collaboration opportunities between academia and industry for the purpose of new material discovery. Our next step is to extend the application to various fields related to creative design and manufacturing.
Establishment time
1st July, 2017