SmartSim: applying deep learning methods to enrich engineering simulations of advanced manufacturing processes

Deep learning is transforming many areas of society, from self-driving cars to cognitive medical assistants and energy efficient buildings. The application of such groundbreaking approaches to the engineering industry has been slow. A primary reason is that deep learning methods are typically developed and implemented by computer scientists and statisticians, whereas, the intimate details of complex manufacturing processes are only understood by engineers. The proposed SmartSim project aims to integrate these fields by developing novel hybrid techniques that combine deep learning methods with advanced engineering simulations. This will drive the development of next-generation manufacturing processes and design techniques.

The manufacturing industry is under increasing pressure to produce components with greater functionality and stricter mechanical, geometric and chemical property requirements. Specification of the multitude of process variables presents the engineer with the daunting optimisation question: “Based on a variety of (often conflicting) performance metrics, which combination of physical, geometric and material variables will produce the optimal component?”. Solving this problem efficiently requires simulation tools, which accurately and efficiently predict process performance and component properties. The current state of the art has made significant progress towards this goal, although widespread industrial application is restricted by two major challenges: accuracy, in terms of predicting material behaviour throughout consecutive processing steps, and time, which limits rapid evaluation of potential designs. The SmartSim project aims to overcome both of these limitations by boosting advanced engineering simulations with deep learning methods in two novel ways: an intelligent constitutive law, and an accelerated solution procedure. This approach paves the way for superior next-generation processing techniques.

The SmartSim project looks to combine machine/deep learning methods finite volume methods in two ways:

• A deep learning elasto-plastic material model
• Accelerated solution algorithms

Funding

This project is funded by the Irish Research Council under the Irish Research Laureate scheme, grant number IRCLA/2017/45.