Simulation, Software and AI | Understanding the current trends in the automotive space
Judy Curran, CTO – Automotive at ANSYS, has 35+ years of experience in the fields of mobility and technology. A former VP of Engineering at Ford, Judy closely works with OEMs and suppliers in her current role, helping them with their digital transformation. In this interview, we learn about the current trends in the automotive space, including physics-based simulation, software-defined vehicles, ADAS, the use of AI in product development, and cybersecurity.
What major shifts have you seen in the product development process throughout your career so far?
The automotive industry has seen a major shift towards faster vehicle development. Traditionally, creating a new vehicle took about four years, but companies aim to cut this down to two years, with some claiming they can do it in just 18 months. As a result, time has become a vital factor, prompting a move towards more virtual engineering and less reliance on physical prototype testing.
As a solution provider, how do you ensure confidence in simulation results, and how well do they align with physical testing? Can you provide examples?
Absolutely. Over time, advancements in computing power—especially with CPUs and GPUs—have significantly increased the speed and accuracy of these simulations. Simulations are based on physics equations grounded in scientific principles, contributing to their reliability. The crucial question is whether you can accurately simulate the product’s actual behaviour. With faster computing power, you can run these simulations at a speed that closely resembles real-time physics.
Simulation capability covers everything from static items like headlamps to complex, challenging scenarios like battery safety testing, marking a substantial transformation in product development.
Examples – With static items like headlamps, optical equations are crucial, as different lamp designs and light sources can create inconsistent light spots, similar to home lighting. Simulation lets you test various lamp technologies in real conditions and immediately see the results, which closely reflect nighttime appearances. Traditionally, this required time-consuming prototype testing in dark rooms. Now, you can run hundreds of iterations in minutes, offering a significant advantage for our customers.
When considering the driver’s experience with reflections and glare, manufacturers previously needed prototypes to test components like chrome trim or textured surfaces. Now, advanced simulations using physics-based equations can model these reflections, allowing designers to optimize elements such as door trims and air vents without physical prototypes. This leads to a better final design before production.
For dynamic scenarios, such as a vehicle crash or a battery thermal runaway, simulations allow us to see each element respond as it would in reality. Thermal events, in particular, are challenging to test physically, but now we can simulate them accurately.
Alongside speeding up product development, what kind of cost savings can simulation lead to?
The savings can be categorized into various areas: the materials needed to build prototypes, the costs associated with building those prototypes, and the personnel involved in testing them. This includes the individuals responsible for the facilities where testing occurs, as well as the expenses related to maintaining and upgrading those facilities.
Traditional methods limit you to testing only four or five design ideas, leading to a cycle of building, testing, and adjusting prototypes. In contrast, simulation allows you to explore hundreds or thousands of design variations, optimizing components for weight reduction and cost savings. This can reveal how using different materials can significantly lower costs. The savings extend beyond just materials; they also encompass engineering efficiency. For instance, one engineer could potentially test a hundred iterations instead of needing two engineers per vehicle in traditional setups.
The real value of simulation lies in potentially leading to unexpected innovations. The industry spends around $157 billion on R&D annually. In contrast, all simulation companies combined likely only account for about $2 to $3 billion. Therefore, when people question whether to invest in simulation, I remind them they spend so much in other areas, making it a worthwhile opportunity. I advise starting with simulations we have already successfully used with multiple customers.
As vehicles increasingly become more software-defined, what areas or functionalities do you see having the maximum impact from the introduction of software?
The growth of vehicle software has been significant. When I started at Ford, I was working on the engine controller, which was essentially the only module that contained software for controlling the combustion engine. Fast-forward 30 or 40 years, and now we have numerous modules handling everything from infotainment to comprehensive powertrain controls, as well as HVAC systems, power windows, doors, and seats. Software is everywhere.
One of the first tasks in transitioning to a software-defined vehicle is to alter the architecture to utilise larger domain computers, zonal computers, or supercomputers. Some new startup OEMs have adopted these larger computing systems from the start. The initial challenge is figuring out how to transition from a distributed system to this more centralized architecture. We provide software that assists companies in evaluating options, such as combining the chassis controller with the engine controller or integrating body and connectivity controls with infotainment.
Once the hardware architecture is defined and the modules are established, the next step is writing software to create new features. When developing these features, testing them on the hardware is crucial. However, the hardware isn’t limited to the electronic components; it also includes the systems being controlled, such as the battery, motor, or camera.
- For instance, in developing an Advanced Driver Assistance System (ADAS) features, we help create models for cameras, radars, and lidars, along with actuators for steering and braking. This allows the OEMs to evaluate the software’s ability to interpret sensor data and send signals to manoeuvre the vehicle.
- In an electric vehicle’s powertrain, we monitor battery temperatures to prevent thermal events while also modeling the electronics and the components being controlled.
- It’s important to track what vehicles were sold in previous years, as maintaining a physical fleet for testing isn’t feasible. If, for example, you decide to modify the parallel parking feature using an over-the-air update, you should be able to test the hardware configurations used in vehicles from the past five years. Therefore, having accurate models of the hardware being monitored and controlled is crucial for successful software-defined vehicle goals.
What role does AI play in the automotive industry today, and how do you see it evolving in the future?
AI presents a huge opportunity. The simplest way to understand this is that the industry frequently repeats certain tasks. For example, aerodynamics – the way air moves over a moving vehicle determines drag, which affects fuel economy and the range of electric vehicles. Every time a new vehicle is designed, any change in aerodynamics, whether it’s the shape of the hood, mirror, or wheels, requires analysis. In the past, we conducted wind tunnel tests, which required a full prototype and took weeks, or we ran full aerodynamic simulations, which would take a week to 10 days at slower computer speeds. With advancements in GPU (Graphic Processing Unit) technology, we can now run aerodynamics simulations in just a couple of hours, but even that is still a long wait.
With AI, we can leverage past simulation data to create a machine-learned (ML) model of aerodynamics for, say, a pickup truck. We can now make modifications in real-time. When designers alter the hood, they can receive feedback from the ML model in 30 seconds.
This is the true benefit of AI. It enhances the processes we perform repeatedly, utilizing vast amounts of simulation and testing data.
The same principle applies to crash testing. Making even slight changes to frame structures can impact safety, and simulating every small adjustment can be time-consuming. Instead, we can use a machine-learned model to expedite the process. Our customers have noted that the accuracy of these models is above 95%, which is not exact but can be close enough when you are investigating many design options.
The strategy is to use these models to explore various designs quickly, identifying those that align best with desired outcomes before running a full simulation on the final design to ensure 100% accuracy.
The benefit applies not only to OEMs but also to suppliers developing components for multiple manufacturers. For instance, if a motor or exhaust supplier serves multiple OEMs in different parts of the world, they accumulate extensive data from previous projects. This data can inform machine learning models, allowing them to generate quicker, more accurate designs and quotes for new projects when approached by an OEM for their next vehicle design.
What’s your read on the current state of the industry? And are there any specific trends you think our readers should keep an eye on?
- One of the major trends I’m observing is delayed launches, which is quite concerning. We’re seeing situations where vehicles are supposed to launch but aren’t ready due to engineering issues. These issues are becoming more common as vehicles grow increasingly complex. I believe that utilizing simulations and conducting more iterations and testing should help address this launch problem that has escalated over the last 4 to 6 years. It’s also about the fact that the complexity of vehicles has reached a point where the traditional methods of engineering simply don’t suffice anymore.
- Currently, there’s a strong emphasis on lowering vehicle costs. The global market has experienced flat sales growth for about five or six years, even with the influx of new OEMs, including numerous EV startups and many Chinese OEMs. While tariffs in the U.S. and Europe aim to restrict the influence of some Chinese OEMs, many industry insiders believe this is not a sustainable long-term strategy. Consequently, there’s a clear push towards developing lower-cost vehicles.
- Another trend to highlight is the increasing interest in India as a potential growth market. Many people are increasingly focused on India, looking to establish their presence and capture market share there. With the infrastructure becoming more developed and consumers becoming more ready for four-wheel vehicles, I anticipate a surge of activity in India.
This interview was first published in EVreporter Nov 2024 magazine.
Subscribe today for free and stay on top of latest developments in EV domain.