Network based systems

Showa Denko develops neural network models to accurately predict mechanical properties of aluminum alloys

Showa Denko KK (SDK; TSE: 4004) has developed neural network models * 1 to predict the mechanical properties of 2000 series aluminum alloys * 2 from their design conditions with high precision in collaboration with the National Institute for Material Science (NIMS) and the University of Tokyo (UTokyo). The models developed allow us to speed up the process to explore optimal compositions and heat treatment conditions for aluminum alloys which can maintain strength at high temperatures and shorten the development time of aluminum alloys to approximately the half to a third of that with the conventional development method, which was not easy in the past.

Aluminum has various applications because it is lighter than iron and easier to work with. However, it is generally used as an aluminum alloy containing copper, magnesium and other additive elements, because pure aluminum has low strength. The development of aluminum alloys capable of maintaining sufficient strength for a particular use at elevated temperatures is desired because conventional aluminum alloys lose strength when their temperature reaches 100 degrees Celsius or more. However, the mechanical properties of aluminum alloys depend on many processing factors, including many types of additive elements and heat processing conditions. The development of high performance aluminum alloys is usually time consuming as the design of aluminum alloys requires knowledge-rich experience of developers and repetition of analyzes and evaluations.

With the aim of solving these problems, SDK participated in a project within the framework of the Council for Science, Technology and Innovation (CSTI), Interministerial Program for the Promotion of Strategic Innovation (SIP), “Integration of materials For a revolutionary design system for structural materials. In this development, SDK, NIMS and UTokyo collaboratively developed a computer system using neural networks, an artificial intelligence (AI) algorithm, to accelerate the development of materials and explore globally the design conditions in aluminum alloy that achieve optimal mechanical properties.

In this development, we focused on 2000 series aluminum alloys, used design data from 410 aluminum alloy records listed in public databases, including the Japan Aluminum Association, and developed neural network models that accurately predict the resistance of aluminum alloys to various temperatures ranging from room temperature to elevated temperature. In addition, we optimized the architecture and parameters of the neural network with Bayesian inference * 3 by applying the Monte Carlo method of replica exchange * 4. As a result, it became possible for us to assess the strength of the aluminum alloy and its uncertainty of prediction. In addition, this neural network model can estimate the strengths of aluminum alloys under 10,000 different conditions in 2 seconds. Thus, it became possible to evaluate aluminum alloys with various design factors exhaustively in a short time.

In addition, we have successfully developed “reverse design tool”, which suggests a set of aluminum alloy design conditions that maximizes the probability of meeting the desired strength at an arbitrary temperature. Thus, it allows us to design high strength aluminum alloys at high temperatures above 200 degrees Celsius.

In its “Long-term Vision for a Newly Integrated Business”, the Showa Denko Group announced that it will continue to be committed to making the most of artificial intelligence and computer science, which are at the heart of its business. basic research activities. We will accelerate our materials development programs by applying the results of this development to our activities to develop various new materials and provide our customers with solutions to their problems, thereby contributing to the prosperity of the company.

The details of the results of this development will be presented during the virtual session of the 2021 Materials Research Society * 5 Fall Meeting, which will be held from December 6 to 8 in the United States and will be broadcast globally via the Internet.

* 1. Neural Network Model: Neural Network Model is a machine learning algorithm that mimics the neural network of the human brain. A typical neural network model has input, hidden, and output layers. The existence of a hidden layer allows a neural network model to learn and estimate the relationships between input and output of complex events. Statistical machine learning with models of neural networks with many hidden layers is called “deep learning”.
* 2. 2000 series aluminum alloys: 2000 series aluminum alloys contain copper and magnesium as additive elements and have high mechanical strength. Duralumin and super duralumin are well-known 2000 series aluminum alloys. 2000 series aluminum alloys are used as materials for aircraft bodies and industrial parts (screws, gears and rivets, etc. .).
* 3. Bayesian inference: Bayesian inference is a method of statistical inference that statistically infers causes from observed facts based on Bayes’ theorem. In this development, we built neural network models that reproduce the correlation between design conditions and mechanical properties of aluminum alloys with Bayesian inference.
* 4. Replica Exchange Monte Carlo Method: The Monte Carlo Replica Exchange Method is one of the computational methods for simulating Bayesian inference with a computer. It is known to converge towards a global minimal solution more quickly than the other methods when solving multimodal problems with many local minimal solutions. This allows us to efficiently explore a wide range of parameter spaces to find the optimal solution.
* 5. Materials Research Society: Materials Research Society is an academic society focused on materials science, established in the United States in 1973. It convenes general meetings twice a year, in the spring and in the fall.

Source link