Do you gamble on digitalisation? A better approach to capitalise on your data
Author : Sam Jonaidi, VP, Enterprise S/W, Automotive & Arturo Vargas, Solutions Mktg Mgr, Transportation | NI
01 April 2022
The automotive value chain is under intense pressure to move towards digitalisation. Organisations are spending billions across a variety of areas, ranging from infrastructure to platforms to multiple applications. To fuel this vast transformation, we must effectively generate, organise, mine & analyse data — the most valuable “currency” we have available — to the highest degree possible.
This article was originally featured as the cover story in the April 2022 issue of EPDT magazine [read the digital issue]. And sign up to receive your own copy each month.
Here, Sam Jonaidi, Vice President, Enterprise Software, Automotive & Arturo Vargas, Solutions Marketing Manager, Transportation at software-centric T&M firm, NI explore why it’s so important that we turn our attention to data to cover the big bets being placed across the industry…
Despite heavy investments in IT (information technology) and OT (operational technology) by the automotive industry, we have witnessed varying degrees of success when it comes to harnessing the full value of data – especially the parametric data that’s generated in assembly, inspection and test processes. Such data is believed to be the richest and most directly indicative of product performance for automotive Tier 1 and OEMs (original equipment manufacturers). So, what explains the mixed results? Two problems are to blame:
1. A lack of focus on making data and its availability as important as functionality when designing workstations, labs, production lines and facilities.
2. The inability to connect data silos intra-company, as well as inter-company.
Prioritising data also supercharges operational performance (OpEx and CapEx), since we can utilise machine learning to preemptively increase yields, reduce scrap and enhance OEE (overall equipment effectiveness) – while OEMs enjoy benefits like full parametric traceability and preemptive recalls.
The value of data
Of the bevy of capital assets that organisations invest in, data is by far the most valuable and ubiquitous of them all. How then can this rich source of untapped insights also be the most neglected? As it turns out, deriving insights from data requires significant infrastructural investment. The Automotive Industry Association in Germany (VDA) projects that EUR 25 billion will be invested in digitalisation by 2024 – and that’s on top of the EUR 50 billion already anticipated to introduce new electric drive systems.
The investment is there; the need is there; and the understanding is there. Yet, something still feels off. Automotive innovators have increasingly understood the importance of using data to gain an advantage over their competitors, but they’ve found themselves frustrated with the lack of solutions that are right for their digital transformation strategy, and for harnessing value to meet their goals.
Figure 1. Variations occur and can signal imminent or future issues; connecting data from different dimensions turns suspicion into actionable insights
The World Economic Forum reported in 2016 that the digital transformation of the automotive industry would yield an estimated US $0.67 trillion of value for automotive players by 2025, with the core motivators being “greater efficiency” and “cost savings.” Five years later, the industry now realises that harnessing data is also key to accelerating reliability, safety and quality. The opportunity isn’t just operational efficiency. We now have a path to leveraging data to improve product performance in this critical moment:
• January 2021—Volkswagen recalled the e-Up due to faulty battery cells and the inherent short circuit or even fire risk.
• February 2021—Hyundai recalled 82,000 electric vehicles (EVs) following reports of battery malfunctions.
• July 2021—General Motors recalled almost 70,000 Bolt EVs due to fire risks.
As the complexity and computational capabilities of vehicles continue to increase, so too does the cost of these related recalls. The Hyundai recall had one of the highest per-vehicle recall costs, of US $11,000. These recalls continue to wreak havoc on a critical industry desperately trying to earn consumer trust in an immature technology. But why?
Because most digitalisation and smart manufacturing endeavors are still focusing on process — not product performance — and analysing data in silos, instead of holistically. These two critical evolutions are at the heart of the frustration with the lack of solutions. By the time you get the data, you’re looking at problems that have already occurred, instead of finding and solving them preemptively. So, what steps can you take to remedy this situation?
Focus on product performance & stop testing backwards
Every system and component in a vehicle passed some sort of test before entering the market. “Passing” meant that, compared to a defined set of specifications, it met enough criteria to move forward in the manufacturing process. The fundamental problem here is that you’re looking backwards by testing something that’s already made. This way of testing focuses on the process, and has become rudimentary when considering the huge complexity of modern automotive systems.
For example, an EV could be experiencing deteriorated range because of bad welding on the battery, software issues on the BMS, a faulty semiconductor, environmental circumstances, driving habits, a combination of those—the list continues. How do we find the root cause for this problem? Traditionally, we would gather all available information and investigate possible causes until we find the source – which could take months or years, cost millions, and possibly, lead to recalls.
Figure 2. Using the welding example, actionable insights are also delivered to the process & equipment, so they can automatically adjust to ensure product performance is consistently delivered
But what if you were to use all your data, at all stages, to the most granular level, to ensure your product will perform and not only pass? What if you could drive actionable insights as you’re doing this? To focus on product performance, we need to understand that when we test for pass/fail, we neglect important data that could reveal issues like variations in the process, diminished product capabilities or short-term failure risk. Most importantly, when a problem does occur, neglecting the product performance data is what forces us to seek the root cause after the fact and end up reworking or recalling.
Since the answer lives in the data, we can proactively test for product performance by connecting data across three relevant dimensions: technology; value chain; and lifecycle.
By connecting all this data from the different systems, across these dimensions, we can activate data analytics techniques to prevent problems, find the root cause on the fly and test for product performance – not only to pass/fail the product.
Connect data across dimensions to break silos
Modern cars comprise massive amounts of technology, from semiconductors to chips, modules, systems, and so on to the full vehicle. Each of these components is typically delivered by different companies within the supply chain, and each company has their own processes, quality checks, and of course, data. These suppliers ship to their customers only their passing products, which may or may not be the best performing ones, and this makes a difference.
Back to the EV example: we know that the quality of the welding in a battery cell impacts the electrical efficiency of the battery pack and consequently, the EV range and performance over its lifetime. The welding involves reliably binding thin, highly conductive, multilayer, dissimilar materials together, and the industry uses multiple joining technologies to maximise conductivity, such as laser welding. After the welding is done in a very precise, repeatable process, EV manufacturers perform a weld integrity test in production by measuring the resistance of the seam and deciding the pass/fail status within ranges of 0.1 mO.
To take such an example from testing for pass/fail to testing for product performance, let’s say we expect the measured resistance of the seam to follow a curve with the centre at 3.2 µO and acceptable variation range within 2.8 µO and 3.8 µO. We perform test in every weld and plot the results on a histogram (Figure 1). If the distribution starts moving, the welds will still pass for some time, but the shift shown by the data will indicate something is happening. Getting the insight before the test starts failing enables engineers to anticipate potential issues, but it’s not enough to start determining the root cause.
Let’s now extend the example and correlate the data to other dimensions, such as a recent change in the supplier of the shielding gas for the welding process, some aluminum oxide building up on the recently replaced probe, or an adjustment on the manufacturing process to speed up the welding. When you’re able to connect all this data on this multidimension data analytics solution, finding root causes, predicting quality and adjusting to meet product performance becomes a reality.
Figure 3. Connecting data across the different dimensions of a product reveals performance data and, with the right analytics tools, actionable insights to eliminate blind spots in the manufacturing process
Deliver actionable insights from the connected data
Connecting the data in the test context includes gathering, sanitising, tracing, linking, contextualising, associating and making it available to the data analytics engine, to then feed the actionable insights back to the right resource. In practice, it means having an eye, a brain and a hand on every parameter measured across the dimensions of the product (and on all of them collectively), and then acting on the insights automatically (Figure 2), as well as informing the stakeholders, so they have zero blind spots.
Turning test from a process that looks backwards into a process that looks at product performance as we manufacture it is an overwhelming challenge that sums up what digital transformation is about – and it requires leveraging data analytics, adaptive manufacturing, artificial intelligence and machine learning.
Despite how daunting this may seem, by doing so, you’re almost immediately able to draw value by connecting data in even one or two of the dimensions. By starting to break down, for example, data silos between the design, manufacturing and in-use product stages, we can make a difference in the value that data brings to our organisations (Figure 3).
You need data insights, not just data
At NI, we understand that companies are at different stages of digitalisation, so flexibility is of utmost importance to us. We have 40 years of expertise generating test data and helping automotive companies use it to ensure reliability, while meeting cost and times constraints.
At our core, we’re engineers who are passionate about enabling innovators to develop products that are more innovative, affordable and safer. Following that passion, the growing need to extract actionable insights drove us to expand from test data to all product data: throughout the product lifecycle, the technology and the supply chain. Through the acquisition of OptimalPlus, we added an established, prescriptive data analytics solution and integrated it across our test technology to help focus automotive efforts on making data-driven decisions and finding root causes, on the fly, so companies can deliver better performing products to market.
We believe that only by being intentional with your data strategy, and through the right analysis of impact and commitment to change, can you deliver the best products to the market. Implementing the NI approach to drive actionable insights from your data and preemptively solve problems is not a gamble on data analytics, but a solid step towards making data work for you to eliminate defects, recalls and waste and meet the fast pace of innovation towards Vision Zero.
Contact Details and Archive...