Creation and Verification of Digital Twins in Cloud Production

TL;DR

This paper proposes using deep neural networks to create adaptable digital twins for polymer composite products, enabling cost-effective design validation and supporting Industry 4.0 cloud production initiatives.

Contribution

It introduces a novel approach employing deep neural networks to develop adaptable digital twins for polymer composite materials, overcoming limitations of traditional physics-based methods.

Findings

Deep neural networks enable digital twins to adapt to environmental changes.

The approach reduces costs by minimizing physical prototyping.

Digital twins improve the accuracy of product characteristic predictions.

Abstract

This article discusses the use of digital twins for products made of polymer composite materials. The design of new products from polymer composite materials, both within the framework of the traditional and new direction of cloud production, requires the need to calculate the physical and mechanical characteristics of the product at the design stage. Carrying out full-scale tests increases greatly the cost and slows down the production. It requires the manufacture of a prototype of the product. The use of existing development tools does not always provide the required characteristics. To solve this problem, it is proposed to use a digital twin, which will not only solve the problem, but will also help to move to cloud production and the development of the Industry 4.0 direction. Thus, a new problem arises - how to create digital twins of products from polymer composite materials. The analysis makes it possible to conclude that the traditional methods of mathematical physics are not suitable for the solution of this problem, since the twins obtained with their help do not have any properties of adaptability. To solve the problem, it is proposed to use deep neural networks one of the most powerful methods of machine learning. This will make it possible to obtain digital twins of products made of polymer composite materials that can adapt to changes in the model, environmental conditions and adapt to changes in the indicators of sensors and transducers installed on the product.