Not so long ago, a doorbell was just a doorbell. Coffee makers didn’t connect to apps and voice-activated assistants, personalize your caffeine fix, and automatically re-order the beans. Many of the things we take for granted in our day-to-day were still merely the work of science fiction.
For almost everyone working in the design and simulation space, complexity is the new normal. At the same time, product development timeframes are shrinking. To add more pressure, most markets are now highly globalized and ultra-competitive. Traditional product design methods aren’t cutting it under new constraints.
Not surprisingly, design and engineering teams are having to rethink their approaches. As time to revenue becomes increasingly critical, so does identifying the optimal product design solution at the earliest opportunity. More conditions and variables need to be assessed quickly, and at the concept stage of product development. Why? Because the later changes are introduced in the process, the more cost and headache they add to the projects.
How designers spend their time is also falling under the spotlight. How many precious hours are laborious pre-processing tasks eating up? How much time are people prepping instead of doing? More than ever, designers need to focus on designing.
In common with just about every other business discipline, the engineering community is looking to data science in engineering for answers. But how can stakeholders leverage this new technology in a reliable, cost-effective way? Does data science in engineering offer teams enough reward to justify spending time on training virtual models? Is the new generation of data science tools accessible to everyone? And how well do these tools fit with the existing simulation-driven and multiphysics environments that are now at the heart of many design workflows?
Here’s the good news. Data science in engineering is already delivering compelling benefits and enabling some radically different ways of working. This goes not just for design and simulation teams, but also in areas such as manufacturing and post-manufacturing, where performance can be assessed for products in operation like consumer goods, aircraft, battery packs, automobiles, bridges, and heavy equipment. Teams can leverage data from such assessments to improve product performance, quality, and safety. If you’re looking for evidence, here are just three of many examples from the data science in engineering toolbox.
Shape Sorting is Your Superpower
When large assemblies are involved, the finite element analysis (FEA) process often requires designers to spend considerable time and effort identifying and grouping similar geometric shapes within a model, a crucial step for ensuring consistent best practices.
But how many precious hours could we save if our tools could learn from our data and instantly reveal geometric similarity for us?
This is where Altair’s artificial intelligence (AI)-powered capabilities are changing the game. In the latest version of Altair® HyperWorks®, AI-augmented shape recognition automates the process of matching shapes, including mirrored parts. Machine learning then enables users to synchronize changes, morphs, and mesh rules. What’s more, the capabilities are seamlessly integrated into the designer’s existing workflows – no specialist skills needed. And best of all, training the necessary datasets is as quick and intuitive as clicking a button.
A New Shortcut for Designers
The typical design method is an iterative process: prepare a model and assess it via simulation. The design team repeats this process until a satisfactory result emerges. This methodical and iterative approach doesn’t lend itself well to acceleration, but with the right tools, a completely different scenario is possible. With the help of AI, users can generate accurate, fully automated physics predictions directly from CAD or mesh data. What’s more, they can produce these predictions up to 1,000 times faster than a conventional solver. That means teams can consider many more options in a short space of time. Only the preferred design needs to be put through a conventional simulation to confirm the AI-generated results. In short, AI-powered physics predictions revolutionizes the design flow. From the outset, everyone can work together at the same time; dozens of design iterations can be assessed side by side, rapidly and accurately.
Integrated directly within the Altair HyperWorks environment, Altair® physicsAI™ empowers designers to leverage previous simulations from any source to make predictions directly from their CAD files. Training the AI model is straightforward and doesn’t require data science expertise. Users simply select the files, variables, and outputs for their study, and physicsAI does the rest. The technology works for any physics, and can handle the complexity of modern multi-system, multiphysics designs. Moreover, models can easily run from a PC or leverage high-performance computing (HPC) in the cloud.
If It’s Not Going to Break, Don’t Fix It
Data science in engineering is also demonstrating its value in the pursuit of greater manufacturing efficiency. Modern supply chains are characterized by just-in-time strategies and lean inventories. Any disruption to production schedules is likely to result in rapidly escalating financial and reputational damage. Traditionally, OEMs have responded by adopting preventative maintenance programs. In theory, these pre-planned work schedules minimize the risk of breakdowns and downtime.
But there are weaknesses in this approach. Production operations must be paused to undertake work that may be unnecessary. What’s more, preventative maintenance is typically slow to respond to changes in metrics such as equipment performance or product quality, which could be early indicators of future problems.
Now, data science in engineering is enabling predictive maintenance. By leveraging data generated by sensors embedded in industrial equipment, along with insight drawn from maintenance records, data analytics toolsets can accurately predict impending failures early accurately. Teams can target repair and maintenance interventions much more efficiently, minimizing costs and downtime. In addition to outright equipment failure, data analytics can also predict performance dips and a system’s remaining useful life (RUL).
Given the massive hype surrounding data science, it’s easy to be skeptical. But data science in engineering is already empowering people to uncover new opportunities by reimagining design, simulation, manufacturing, maintenance, and end-of-life strategies. Altair’s flexible, open data science tools are being integrated seamlessly into existing workflows, ensuring they’re as accessible as they are powerful. It may not quite match Hollywood’s idea of a superhero, but data science now has the power to speed the journey to brilliant engineering outcomes, and help enterprises negotiate their way through an ever-more complex universe – and that’s something of a superpower itself.