Has remote working during COVID-19 changed the product development process forever?

Has remote working during COVID-19 changed the product development process forever?
15 September 2020

In this Q&A, James Mullineux, Ricardo Head of Digital Engineering, discusses digital engineering, simulation and analysis and considers whether remote working during COVID-19 might have changed the product development process forever
Q: How have simulation tools for powertrain development advanced in recent years?
James Mullineux: Five years ago, while the electrified powertrain was certainly around, now, everything is electrified so very early on in concept, you cannot decide what engine strategy you want without also thinking about what hybrid strategy you’re going for, what battery size, what emachine, and what transmission technologies you are going to put into the box.
Back then you could be a little bit more isolated in terms of decisions. You could choose an engine concept almost independently of your emachine concept, and this was reflected in the available simulation tools: you'd have a very strong engine biased simulation suite and probably a simulation toolset developed over many years that dealt with that pretty robustly.
All that is changing now, and I would say that there are five key areas where simulation tools have advanced:
One of the areas we are focusing on now is the appropriate combination of simulation tools to model systems. As powertrains have changed with the rise of hybrid, but alongside engines, we need a toolset that can model all propulsion elements at the appropriate fidelity.
We are seeing the need for significant elements of electrification in most of the industry sectors that we are covering. Being able to incorporate electric machine models, battery models, transmission models, engine models and thermal models appropriately is probably one of the key areas.
If you think of a classic V diagram, then we are looking at how we use simulation both to drive development activity down the left hand side where we are creating, defining and scoping the product as well as increasingly virtualising the validation activities on the right hand side, particularly calibration.
Another key element is making sure that it is not just a one size fits all: different validation activities will want different simulation models. It’s not just a case of having a single modelling set, but considering what is the problem that we are trying to answer, and what is the validation that we need to demonstrate in choosing the set of models that are most appropriate to answer that question.
One of the big areas that we are trying to look at is virtual calibration and testing particularly around some of the critical attributes performance emissions, energy consumption, drivability and so on. Traditionally, an approach by some in the industry has been to offer OEMs a virtual testbed, a physical asset, operated and controlled by a calibration engineer, that has a testbed controller and a HiL rig, into which will be placed very particular models that represent a virtual testbed. Traditionally, this has tied an OEM to a physical asset which, as the ongoing global pandemic has demonstrated, can be extremely limiting, and additionally, the technology that goes into those simulation models is always the same: a standardised engine model, or transmissions model and so on. So when you then want to look at something that’s more advanced or different, you can’t change it because the modelling method is constrained.
Contrary to the ideal, there is no single solution in terms of simulation where we would just say here is solution tool x and it will just model everything you need in all the fidelity you want and that’s all you’ll ever need to deal with. The uncomfortable truth is that it is a case of assembling elements which will answer specific questions. For us, the biggest change is about offering scalable or appropriate fidelity based on the attribute. We ask: what is the challenge we are trying to solve, and how do we solve that in the most efficient way, trying to avoid being tied to a physical asset.

Q: What can now be done virtually and how does this change the development process?
James Mullineux: This is a continuum. What we are particularly looking at a lot is virtual calibration and around shorter timescales, prototype reduction and therefore cost reduction. That’s particularly prevalent in products that need to work globally and which have a reasonably high number of variants and adaptations to suit different markets and different needs. Physically validating these is challenging because of the need for large numbers of all different prototypes; the ability to virtualise those changes and assess risk from change is really important.

The other element that has come in is real driving emissions (RDE) where it is no longer a single legislative drive cycle for which you can work to develop a robust solution. It’s now a far broader category of testing where it’s not possible to test every valid real driving cycle. It becomes a statistically driven methodology, using RDE cycle generators, running all of those through simulation models to identify worst cases, and then those are the cases that you can test physically to confirm that the product is robust. This is one of the elements where we see emissions legislation driving changes to the development process in particular, relying on the virtual simulation toolset.
More powerful computers and access to clusters for modelling have enabled large scale deployments of statistical analysis on driving emissions and this is probably the area that has transitioned. In addition, there are constantly developing new physical and empirical models that allow us to have the appropriate fidelity level to answer the questions that we need to answer.
Q: What are the hardest aspects to simulate, and where does simulation technology need to expand its capabilities?
James Mullineux: We tend to find it is where chemical processes become important: emissions from combustion engines, battery degradation and getting that right and then after-treatment performance.
Historically, some of those methods have relied on quite detailed analytical models. We take 3D CFD for combustion modelling, which is looking to try and predict emissions, and even that can be a challenge for certain emissions.
For virtual calibration, which requires models that run around real time, we can't rely on those detailed methods, and we need something much faster. So, then you start to bring in elements of empirical and maybe categorizing your combustion system or your after-treatment system or your cell degradation.
That’s alright at one level: you can do component tests or subsystem tests to understand how a cell performs, get degradation of the cell, and then map that out across the battery pack and then across into a vehicle.
However, the ability to predict the cell degradation is really challenging. For some tasks that's fine, because the cell development is sufficiently advanced from the vehicle developments that generally, you’re not using very cutting edge cells where the cell chemistry is changing at the same time you are trying to develop the vehicle. A lot of OEMs will buy in cells which have already been developed enough where you can test them and characterize them. There are areas of continuous learning in how battery packs then operate in real world usage cycles meaning that you may want to change battery management systems or make other changes to the way energy is being used around the vehicle.
The chemical processes associated with say emissions or after-treatment rise and fall depending on where the area of focus is. With the passenger car switch from diesel, the focus on modelling also shifts from diesel emissions to gasoline emissions. Some of those are cleaned up by after-treatment technology very robustly and therefore the level of simulation needed is perhaps reduced because we know that it works but then you’re constantly trying to reduce development effort and the cost of those components, so it's the hardest point when you're on a fine threshold of something working and therefore the accuracy of the model is really critical on passing or failing.

Q: Has COVID-19 remote working changed the development process forever?
James Mullineux: Ricardo employed a ‘digital first’ strategy, which saw the creation of 2,000 home offices as employees worked safely from home, and we were able to continue to deliver services to our global customer base and products safely but remotely. Generally, it was business as usual. We still relied on CAD and some of our 3D CFD areas on relatively high powered machines for manipulating geometry, be that that in CAD world or in some of the analysis world. We just had to choose where to put the slightly more powerful desktop machines, but most of our solving is already centralized. We’re still solving in clusters here at our Shoreham Technical Centre or at our technical centres in Leamington, Cambridge and Prague, so this sort of communication is something we are very familiar with.
One of the elements we are looking at is the transition from desktop-based activity where an individual engineer has his task on his desktop through to the collaborative environment. In effect, how to transition from desktop to server based or Cloud based methodologies. We did that with some of our CAD for example, and some of our analysis. However, there are still pockets where it is fairly segmented into single activities locally. We, of course, very quickly got used to remote working for teams and communication and documents. Transitioning that for a whole variety of tasks is really interesting from a simulation and analysis point of view.
Adopting purely virtual methodologies means that you’re not constrained by again needing to be attached to a physical piece of equipment. Traditional service providers in the industry undertake calibration using a physical asset test bed.  There may be a way now of connecting to it remotely, but it’s a physical ‘thing’ that has to be located somewhere and which requires someone physically to sit at it to operate it. Not constraining yourself to a physical asset is a really important element.
If we can transition into more Cloud based methodologies of appropriate fidelity, then that’s completely unconstrained by where people are sitting physically. Additionally, you can add to that elements of automation where things are just running in servers where you check in some new change to a model, or to calibration, and it goes off and runs the test suite of conditions to verify that that change is acceptable. We will continue to see this, but probably the concept of remote working has accelerated the recognition that collaboration needs to work effectively.

Q: What are the biggest challenges for engine development over the next five years, and how can simulation technology help overcome them?

James Mullineux: Thinking about potential long-term trends which might result from the current global pandemic, there have been reports of cities across the UK and mainland Europe considering how they might be able to retain the air quality benefits which have been the result of greatly reduced urban vehicle traffic during the lockdown, through a permanent reduction or even a ban on no private transport in city centres. We have also seen an exponential rise in online shopping and take away deliveries. What these trends could mean for OEMs is: a move away from the traditional focus on passenger car production, towards something more heavily biased towards light commercial vehicle fleets where OEMs want to spend their money; a greater focus on mobility solutions for centralized or city based transport; and a shift in focus to intercity travel or longer distance travel as the main driver for private vehicles. This is really interesting in terms of what challenge that puts into the demands for different vehicle types and the right powertrain for the future.
In many ways, the current pandemic is the most extreme example of the elements of uncertainty and complex challenges for OEMs as they have to pursue relentlessly product efficiency while driving down tailpipe emissions, whether that is through non-carbon based fuels, like hydrogen or through after treatment solutions and other technologies which will reduce the tailpipe impact.
The key challenge in all contexts is: what is the right propulsion system for a huge variety of applications. Cutting through the uncertainty to find the right answer to that question, and perhaps revisiting product planning as a result, is something that simulation is really designed for. Simulation tests a wide range of scenarios with matrices of technologies: engine technologies, emachines technology, battery technology, transmission technologies to draw out for that application what the best combinations of technologies are for particular purposes, with particular types of attributes.  This is where simulation is really powerful as a planning and strategizing tool, and then you can make decisions on cost and attribute trade-offs without needing to build things.

Q: Could virtual validation become the norm?
James Mullineux: Validation has a few elements. One is durability and reliability, for which elements of physical testing will remain because there are some complex elements involved. Data analytics are being used somewhat now, but we will really start to see the growth in their usage, and real world data analysis to understand the relationship between in-use performance and how products are developed. This includes how products are wearing out or failing and feeding that through into the development strategy so that you can again minimize any testing time by relying very heavily on the data you have got in the real world.
The other area is working with the regulators on appropriate virtual validation methods. It could be that there is a transition towards validating the methodology and the process to produce a product rather than a single final test that simply confirms that a product has passed tests. Until legislative bodies accept a virtual method as legitimate, it won’t be possible to validate products virtually.
To find out how Ricardo’s simulation and analysis capability can help accelerate your product plans thorough reduced timescales, costs and prototypes, and improved adaptability to diverse global regulations, please contact: [email protected]