HPC simulation: a resourceful route from the drawing board to the driveway

High-performance computing boosts the synergy between simulation and manufacturing.

High-performance computer (HPC) simulation using computer-aided engineering (CAE) applications has evolved from a tool once used by automakers for limited-use cases in the vehicle design process to improving and optimising all aspects of product development, including the manufacturing process. It is now common for the same CAE simulation tools to analyse and determine the most efficient and cost-effective manufacturing processes.

Widely used for sheet metal stamping simulations, HPC simulation removes inefficiencies from the eventual manufacturing process by preventing potential defects such as metal wrinkles. Simulation has become especially valuable as new materials such as aluminium and plastics are introduced, enabling manufacturers to avoid the expensive inefficiencies of the traditional trial and error approach to establishing the production process of a new vehicle.

As new materials are implemented for the advancement of vehicle design, manufacturers can no longer depend on the experience and expertise that they traditionally relied upon to understand the extents to which these new materials can be manipulated. HPC simulation is essential in enabling manufacturers to understand what can be achieved with these materials and can predict the final shape of a new component through the digital replication of material properties.

Driving up the data flow

Throughout the automotive industry there is growing demand for data gathered during manufacturing simulation that can be used to improve the finalised design of a vehicle.

To achieve a higher fidelity in results, it is no longer adequate to use nominal or average material properties for individual components. The structural properties and tolerances of the material will change as the material is manipulated throughout the vehicle’s production, therefore it is more accurate to use the final material properties of the component after the manufacturing process.

For example, the simulation results of a piece of sheet metal being bent into a certain shape, such as a body panel, can provide the material’s stress and strain state without a physical piece of metal having to be bent. With high-fidelity crash simulation, the slight difference in material properties before and after stamping will affect the crash response, so including this effect will deliver a more accurate simulation.

The results speak for themselves, and so there is now a growing requirement for increased communication between teams working on different stages of the manufacturing process and the flow of data from the manufacturing process back into the vehicle CAE analysis process.

The stochastic approach for a robust design

Accounting for the variation in the real-world structure after manufacturing has led to the use of a ‘stochastic approach’ in CAE simulation. In other words, a robust design approach that ensures the final product is highly resilient against slight variations that result from the manufacturing process.

A good example of this is the location of spot welds in the vehicle’s construction, commonly used to connect large components in the structure, such as the floor pan and the body shell. It has been shown that even the slightest variation – including variations within manufacturing tolerances – in the location of a spot weld can have a major impact on the safety performance of the finished vehicle.

To understand the sensitivity of the design, a stochastic simulation process runs hundreds of simulations, which vary the parameter of interest. Comparing the results allows the manufacturer to determine the sensitivity of the design to the spot weld location or other parameters. Understanding this sensitivity enables a more robust vehicle design, which reduces the risk of failing a safety test due to the slight variations in manufacturing.

This robust design does require an investment in HPC compute power. The stochastic investigation of a single component in a vehicle’s design requires hundreds of simulations and, for each of these simulations to run simultaneously, a greater amount of computing is required. The computing requirements have traditionally been a limiting factor in adopting the stochastic approach but the increased availability in HPC power now makes this practical for automobile design and production processes to run more efficiently.

Reinventing the wheelbase

Further development lays in the synergy between manufacturing technology and CAE simulations – the results of which advance additive manufacturing technology and the use of topology optimisation tools. For example, topological optimisation might generate a more organically shaped vehicle component that could be shaped similarly to a human bone to add more strength and resilience, while reducing the weight of a traditional structural component such as an I-beam.

Topology optimisation has been available for many years, but in the past, this approach required a huge amount of computing power and the final ‘optimised’ design would have a very complex shape that would be impractical to manufacture.

Today, additive manufacturing processes make it reasonable to produce these very complex parts and the increase in computing power makes the amount of simulations required far easier to achieve. The combination of these technology advancements creates an entirely new set of design possibilities and will play a key role in industry initiatives such as vehicle weight reduction.

Souped-up data performance

Not only are we seeing a huge increase in the amount of data available but also in the requirement to share this data throughout the organisation. Simulation is producing high-fidelity results for a wide-range of design options while manufacturing process and tools are producing highly valuable data. In some cases, data is also being streamed back to the manufacturer via the Internet of Things (IoT) once the vehicle is on the road.

Incorporating sensors that can relay information back to manufacturers contributes to the wealth of information on the past, current and potentially the future performance of the vehicle, or similar models and conceptual designs based on the same vehicle platform.

Clearly there is a tremendous amount of value in this data and the ability to analyse this, finding the patterns and gaining actionable information, is an aim every automaker wishes to fulfil. The challenge lies in recognising what to look for and how to find this value in the enormity and variety of data available.

Just like supercomputing power has enabled the simulation of the vehicle design and manufacturing, that same power can be applied to the data analytics challenge and is opening up a new field of high-performance data analytics (HPDA).

Although still a relatively new field, automotive companies are making huge investments in data analytics. There is no shortage of potential use cases, from spotting warranty or recall issues earlier, to optimising the manufacturing processes, developing better understanding of industry trends and learning about evolutions in consumer requirements.

These are not necessarily new concepts but the technology available is evolving rapidly. While implementation in production infrastructure remains a work in progress, the uptake throughout the automotive industry is likely to gather pace sooner rather than later. Much like the increased adoption of vehicle electronics, autonomous driving and other industry trends, HPDA will redefine the technology and the associated skills required to compete in the automotive industry.

Greg Clifford is the manufacturing segment leader at CRAY.

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them