2020 Volume 61 Issue 11 Pages 2041-2046
Materials play key roles to solve social problems. In order to accelerate materials research and development, it is crucial to use the power of cyberspace. This article overviews the concept and current status of Materials Integration (MI), a new concept that was proposed in the first-term Cross-ministerial Strategic Innovation Promotion Program (SIP) “Structural Materials for Innovation” and that has been centered in the second-term SIP “Materials Integration for Revolutionary Design System of Structural Materials.” The concept of MI is to computationally link the material four elements of process, structure, property, and performance for replacing experimental trials and errors in physical space by computational ones in cyberspace. The MI can be characterized by systems approaches and the deep use of data science. The first-term SIP demonstrated the proof of concept for some example problems by developing a computer system containing of computational modules and workflows connecting them. In the second-term SIP, the MI system has been further developed to solve inverse problems, i.e. to design materials and process from a target performance. Moreover, the target materials and processes are expanding to advanced ones used in aerospace and power generation industries as well. The article discusses the outlook of the MI-system based platform for accelerating materials innovation by academia-industry collaboration.
This Paper was Originally Published in Japanese in Materia Japan 58 (2019) 489–493.
Fig. 5 Future outlook as a platform for academia-industry collaboration.
The role of materials in solving social problems is increasing more than ever. In the energy and environment fields, for instance, innovative structural materials that are light, strong, and heat resistant are required to improve the energy efficiency of transportation and power generation equipment and to reduce CO2 emissions. The performance requirements are becoming stricter year by year. Structural materials that are used in a harsh environment for a long time need experimental verification to ensure reliability, and development is a slow process. The difficulty of controlling the microstructure, which is mostly in non-equilibrium state, contributes to slowing down research and development (R&D). To respond quickly to societal demands, achievement of a new, faster R&D approach is vital. R&D can be accelerated by enrichment and dissemination of computational materials engineering and incorporation of advances in cyberspace technology such as recent drastic enhancement of computer performance and rapid development of data science.
Based on this recognition, a concept of Materials Integration (MI) that accelerates materials development by fusing cyber and physical approaches is attracting a lot of attention; for example, it was adopted as a subsystem that supports Society 5.0 by the Council for Science, Technology and Innovation of the Cabinet Office (see Refs. 1, 2) on the integrated materials development system). The MI’s concept had originally been proposed in the Cabinet Office’s Cross-ministerial Strategic Innovation Promotion Program (SIP) “Structural Materials for Innovation,” and the basic, computational system of MI was systematically developed. “Materials Integration for Revolutionary Design System of Structural Materials” (hereafter “Materials Revolution”), where the main focus is MI for solving inverse problems, was adopted in the second phase of the SIP and research that would lead to a novel materials R&D approach has begun.
This article summarizes the philosophy of MI. It provides an overview of the achievements in the first phase of the SIP and the aims of research in the second phase of the SIP. In the light of the global situation, we propose that the MI system being developed should be a platform for academia-industry collaboration.
A key process behind materials R&D is to make a link among the four elements of process, structure, properties, and performance. MI is a concept to link the four elements of materials on a computer by using experimental data, calculation, theory, and data science for thorough predicting performance from process in order to accelerate materials development. We understand that “Integration” in MI is meant to encompass all useful engineering methods, even those for which the theoretical mechanism is not yet fully understood. For example, suppose that one attempts to predict creep performance (lifetime and location of damage) on a heat-resistant part containing a weld zone. Though the process of welding is far from being fully understood, the thermal history of the base material can be calculated by thermal conduction analysis using a phenomenological model. Appropriate parameters for the phenomenological model can be estimated from temperature measurement and observation of the macroscopic structure at the weld zone. The tuning of the parameters using experiments yields a precise prediction of the macroscopic morphology of the thermally affected zone with reasonable accuracy. Similarly, one can calculate the creep damage of the entire part by numerical simulation, using previously proposed models on creep deformation and damage together with reasonable parameters identified from creep database.3) Combining these as in Fig. 1 makes possible the prediction of creep damage in the entire part from the welding conditions through estimation of the macroscopic morphology of the thermally affected zone. Every element from process to performance can be linked by flexibly integrating various methods: for instance, by integrating numerical simulations of welding and creep damage with data science methods to extract model parameters from the database. This is the philosophy behind MI.
Workflow to predict creep damage performance of a heat-resistant structural part containing a weld zone.
The characteristics of MI are the adoption of systems approach and the use of data science that can improve the effectiveness of prediction. The systems approach provides perspective on what areas urgently require fundamental investigation to develop the targeted material. This is useful to realize an organic collaboration between the sectors of basic research (universities and national institutes) and of applied research for making materials practical (industry). Use of data science, which is the second characteristic, allows effective and rational applications of data from precious and expensive experiments.
MI was designated as one of four research domains in the SIP “Structural Materials for Innovation”, which was carried out over five years starting in fiscal year 2014 (see the JST website4) for details). Figure 2 shows an outline of the development of the MI system. Fourteen organizations in academia and industry participated, with major contributions from the University of Tokyo and National Institute for Materials Science (NIMS).5) A computer system dedicated for materials development was designed and implemented, based on the hypothesis that to achieve MI in a usable form, it is necessary to have a scheme that implements various methods as flexibly connected modules on a computer. The key targets selected were to predict microstructure and time-dependent performance, including fatigue and creep, since they are crucial for structural materials. With the addition of data science usage, we organized the following four topics and units: microstructure prediction, performance prediction, data assimilation and machine learning (using data science), and integrated system (Fig. 2).
Outline of the MI system development project in the first-term SIP “Structural Materials for Innovation”.
Fatigue, creep, brittle fracture, and hydrogen embrittlement of the weld zone of steel were designated as example topics for development, and the obtained results were deployed to aluminum and titanium alloys. Steel is one of the most difficult targets because its performance is extracted by highly sophisticated microstructure control including various types of phase transformations and precipitations. Therefore, a scheme that can handle steel is expected to be applicable to a wide range of materials. The welding process contains solidification as well as diffusional and diffusionless phase transformations; thus a framework that can treat welding can treat thermal treatment processes of metallic materials in general. The goal of development in the first-term SIP was to demonstrate proof of concept using specific examples with versatile applicability in mind.
The Materials Integrated system (named “Mint system”) was completed and a scheme to connect modules flexibly was established after approximately four and a half years of development. More than 120 modules for structure and performance prediction were developed, and more than 100 workflows connecting these modules were implemented in the Mint system. Important data science methods, such as data assimilation and model selection, were developed, and the achievements were installed as libraries that can be used from the Mint system. Various published datasheets, mainly on steel materials, were collected and digitized, and a database containing more than 20,000 data points was compiled. These achievements demonstrated that it is possible to thoroughly predict performance from process with the example topics, and proof of concept was demonstrated with examples, and proof of concept of MI was accomplished. The achievements in each topic are overviewed in the present special issues and a part of them have been published in a special issue.3,6–12)
After completion of the Mint system as version 1.0 and achievement of proof of concept of MI, advancement to the stage of the project toward social implementation became possible. Social implementation in this context means that MI is actively used in industry R&D; the program envisions a future where MI assists to develop many revolutionary materials that drastically change society. Based on this background, “Materials Revolution” was adopted as the second-term SIP. The research targets and program director were approved in November 2018, and the research started with an intended duration of five years (until March 2023).13)
Application to inverse problems and extension to advanced structural materials and processes are the keys to social implementation. The former case is elaborated first. The aim of R&D in industry is to find new materials and processes that can meet performance requirements that are set by societal demands (materials users). Hence, they are always solving inverse problems. In other words, they need tools to deduce the optimum materials and processes that meet predetermined performance goals. Therefore, for MI to be utilized in everyday R&D in industry, it is necessary to demonstrate that MI is useful in solving the inverse problems on which researchers are working. Development of inverse problem MI was set as a task based on this recognition.
Figure 3 shows the approach taken to inverse problem MI. Traditionally, a hypothesis is experimentally verified and due to the high cost of experiments only a limited number of trial and error could be conducted on a limited number of variables. Although the new MI-based development approach still takes trial and error to be a crucial step, it relies more on computer calculations instead of laboratory experiments. Hence, trial-and-error-associated cost is reduced drastically. Moreover, we can rapidly discover overlooked potentially optimum materials and processes through application of data science, thereby making trial and error more efficient and comprehensive.
Idea of inverse problem MI in the second-term SIP.
Another factor critical for social implementation is to increase the number of applicable materials and processes. The Mint system itself is designed to be versatile. However, to solve actual problems concerning various materials and processes, development of modules and workflows to handle the target is necessary. As targeted advanced structural materials and processes we have chosen those of interest to aerospace and electrical power industries in this project from the two points of view: one is that Japan leads the development of materials for airplane and power generation applications; and the other is that the R&D time tends to be longer and longer in this field. Application of powder processes is starting to be implemented in airplane engines and power generation turbines; thus, we have been developing inverse problem MI for powder processes in heat-resistant metallic materials (specifically, three-dimensional additive manufacturing processes and powder-metallurgy or -forging processes). It should be possible to handle these processes by efficiently utilizing the numerous modules developed for steel in the first term. In addition, we apply the concept of MI to carbon fiber reinforced plastics (CFRP), in which Japan has a strong advantage. We will build a set of multiscale computational tools useful for development of CFRP by leveraging seed technologies including quantum chemistry calculations, mesoscale calculations on phase separation of polymers, and prediction technology for mechanical properties of structures containing fibers.
We point out four issues on inverse problem MI design (Fig. 3). First, it is important to create inverse-problem examples to which data science is applied. We must establish a forward-problem-calculation workflow for predicting performance from material and process as soon as possible, and then will be able to apply various methods in data science for optimizing the materials and process to obtain the target performance. Next, building detailed models with high precision is important. Factors that are not considered cannot be optimized, so modeling that includes as many factors as possible is desirable. Third, the Mint system needs to be improved to automate optimization calculations. Finally, schemes whereby companies can internally use their proprietary data are important. Designing a standard data architecture to describe the data necessary to solve an inverse problem would promote usage of proprietary data within companies.
The “Materials Revolution” SIP consists of three domains: domain A that develops inverse problem MI, domain B that deploys it to CFRP, and domain C that deploys it to heat resistant metals and ceramics.14) Domain A is a collaboration of 28 organizations in industry and academia with NIMS and the University of Tokyo at the core.
The attempt to develop materials in cyberspace is global. Figure 4 summarizes worldwide efforts in this direction, showing various international attempts to solve multiscale, multi-physics problems that occurs in actual materials R&D. The Mint system developed in the SIP handles various computational modules and connects them to deal with complex materials problems. Similar systems are found in the Center for Predictive Integrated Structural Materials Science (PRISMS) at the University of Michigan in the US;15) the Interdisciplinary Centre for Advanced Materials Simulation (ICAMS) at the Ruhr-Universität Bochum, Germany;16) QuesTeK, a venture company in the US;17) and ProperTune,18) developed in the Finnish research institute VTT. Connections in PRISMS and ICAMS are limited to a number of already available workflows, thus versatile connections are not intended. Therefore, connection of modules must be done manually. Access by general users is not expected in QuesTek and ProperTune; hence, the actual mechanism by which multiple modules are connected is unknown. However, these tools were developed primarily for in-house use to make a solution, meeting the demand for outsourcing of materials R&D. Consequently, there would seem to be weak motivation to invest in developing a system that achieves automatic connection; we speculate that the developers manually connect each computational tool as necessary. The European Materials Modeling Council19) is a European organization aiming to standardize file formats and simplify transfer of data between modules. However, generalization is difficult, because the type of data being handled varies with each materials problem being addressed; hence, standardization is not progressing steadily. One strategy (other than that of connecting modules) is to use a specific solver at the core and then connect subroutines to treat complex materials problems. Macroscopic mechanics calculations using finite element analysis are chosen as the core, and mesoscale procedures that calculate the microstructure are incorporated as subroutines. This is an excellent approach to solving a specific problem, and efficient development and calculations are possible. There are examples to predict material properties in forging processes. However, the problems that can be solved are limited by the calculation method at the core, so the approach is not very versatile. This overview finds that the Mint system is unique in the world because it has a method to connect modules flexibly and automatically.
International comparison of computational materials science approaches to real materials problems.
MI is not only a tool to accelerate materials development but is also expected to function as a platform for academia-industry collaboration. Figure 5 shows the outlook. Needs-driven R&D by both of industry and academia are important to achieve innovation. Development toward commercialization will be accelerated by letting companies use the achievements of the academia-industry collaboration as modules and workflows in the Mint system. Meanwhile, the Mint system could also be useful for making seeds-driven research practical. For instance, when a university or national institute makes a new discovery on a material, companies can quickly begin conducting applied research in cyberspace by implementing a module based on this discovery on the Mint system. Accumulating R&D achievements as modules and workflows increases the range of targets that the Mint system can handle, and the quality of solutions will improve. Thus, the Mint system can be considered as a platform that raises the standard of Japanese materials R&D by compiling centrally in cyberspace outcomes from each research. Therefore, urgent discussion is necessary to set up an organizational framework for how the Mint system should be utilized.
Future outlook as a platform for academia-industry collaboration.
This article provided an overview of the current situation of MI as developed in the SIP. A proof of concept was demonstrated in the first-term SIP, and development toward social implementation is progressing in the second-term SIP. The Mint system developed in the SIP is unique in the world and is expected to develop as a platform for academia-industry collaboration. R&D using MI is expected to spread beyond the SIP, and we hope that this article will be helpful in achieving this goal.
This research was conducted by Cross-ministerial Strategic Innovation Promotion Programs (SIP) “Structural Materials for Innovation” and “‘Materials Integration’ for Revolutionary Design System of Structural Materials” (managed by the Japan Science and Technology Agency (JST)) by the Council for Science, Technology and Innovation of the Cabinet Office. We would like to express our gratitude here.
