Advanced Computing in the Age of AI|Friday, May 29, 2020
  • Subscribe to EnterpriseAI Weekly Updates:  Subscribe by email

Digital Manufacturing: Why There’s Never Been a Better Time 

The past several decades have seen the continued outsourcing of manufacturing to overseas outfits. Increased competition, both globally and at home, is leading US manufacturers to seek out all available avenues to cut costs and boost efficiency. With the cost of entry to the world of digital manufacturing at an all time low, there has never been a better time to make the switch.

The past several decades have seen the continued outsourcing of manufacturing to overseas outfits. Increased competition, both globally and at home, is leading US manufacturers to seek out all available avenues to cut costs and boost efficiency. While plenty of big-name manufacturers have already made the leap to a software-based workflow, the community's so-called "missing middle" has been more hesitant in adopting advanced computing technologies. It's not that they don't want to: Out of the approximately 300,000 small and medium-sized manufacturers (SMMs) in the US, 80 percent would use modeling, simulation and analysis tools if they could (Source: NCMS/Intersect360). But a variety of barriers — infrastructure, cost and talent concerns — impede the process.

flickr photo by Flor MBut are these actual obstacles just perceived barriers? There's a growing awareness of the need for digital manufacturing tools, as highlighted here, here and here, but what may not be so clear is that the computational and software resources themselves are now more accessible than ever. Multiple IT trends have culminated in a veritable trifecta of opportunity. The falling price of processing power, the performance-enhancing effects of graphics processing units (GPUs) and a wealth of cloud computing solutions mean that there has never been a better time to make the switch.

The democratization of HPC

The cutting-edge of computing is a moving target. With non-stop innovation at the high-end, resources that were once available only to the most demanding user eventually become accessible to the mainstream business user and then to the average consumer. The current generation of multicore desktops can run applications that were once the exclusive purview of expensive clusters. The trend, often referred to as "the democratization of HPC," was the subject of a presentation given by Stephen R. Wheat, Intel's senior director of HPC worldwide business operations, at the 2011 HPCC Conference in Newport, RI. Wheat explains how the top-down innovation flow has placed previously inaccessible technology into the hands of the missing middle who then benefit from the competitive advantages it confers.

The falling price of hardware is one of the hallmarks of the democratization of HPC. According to Steve Conway, research vice president in IDC's high performance computing group, a supercomputer that cost a little over $5 million in the 1980s would go for around $1 million in the 1990s. Today a system of equivalent computing power can be had for around $100,000. That's a decrease of 5,000 percent! IDC studies cited by Conway list the challenges SMMs face in the move to advanced computing, such as upfront costs and talent considerations, but companies that do make the leap — and here Conway names aerospace/automotive suppliers BMI, Intelligent Light, L&L Products and Swift Engineering — report increased innovation and faster time-to-market. Organizations such as the Council on Competitiveness, the National Association of Manufacturers, the National Center for Manufacturing Services, and the National Center for Supercomputing Applications are a few of the key players that are working to ease the transition to digital manufacturing.

More for less with GPUs

GPUs, popularized by chipmakers NVIDIA and AMD, are made up of lots of parallel processors, which makes them ideal for highly-parallel processing tasks like simulation and modeling. They also offer a lot of processing power at a relatively low cost, and as such are becoming nearly ubiquitous in cluster and supercomputing systems. Depending on the actual algorithms being run, GPUs can boost performance by an average of 10x and sometimes by a factor of 50x or more. According to company literature, NVIDIA's Tesla 20-series GPUs deliver the same performance of a traditional CPU-based cluster at one-tenth the cost, using one-twentieth the power. In perhaps the ultimate portrayal of GPU might, China used the graphics chips to takeover the top spot on the TOP500 list of world's fastest supercomputers.

GPU-accelerated software is suitable for mechanical design applications in virtually every manufacturing-based industry, including aerospace, automotive, energy, electronics, life sciences, industrial equipment and consumer goods. Software modeling vendors such as Daussault Systèmes, ANSYS and ACUSIM Software are all leveraging the power of GPUs to reduce the processing times of their design software. Just last month Dassault Systèmes announced that its finite element analysis (FEA) product suite Abaqus 6.11 was able to run computer-aided engineering (CAE) simulations twice as fast using NVIDIA's Tesla 20-series GPUs as with CPUs alone. ANSYS and ACUSIM also demonstrated a doubling of performance with the Tesla chips, compared with the latest quad-core CPU running the same simulation.

Cloud computing levels the playing field

As manufacturers begin employing digital models to virtually test their products or to run other segments of their business, they'll need additional compute power. They have the option of purchasing those resources outright or they can rent them on an as-needed, pay-per-use basis. The second scenario falls under the heading of cloud computing and can offer considerable savings over an in-house system. While there are a gamut of issues that arise when considering a move to the cloud — among them legal, security and standardization concerns — the field is actively evolving to meet the needs of businesses both small and large.

Solutions from Microsoft and Autodesk illustrate the diversity of cloud computing options. Microsoft cemented its cloud commitment in early 2010 when the Windows Azure platform opened for business. With a pay-as-you-go model, applications run on Microsoft's datacenters instead of in-house servers and scale during times of peak demand. In April of this year, Microsoft announced its Reference Architecture Framework for Discrete Manufacturers Initiative (DIRA Framework) aimed at driving cloud computing solutions across manufacturing networks and tightening collaboration among value chain partners. In February, software vendor Autodesk expanded AutoCAD WS, which uses cloud computing technology to enable AutoCAD users to view, edit and share their designs through Web-based devices. Autodesk's other cloud-based applications include an energy analysis tool called Green Building Studio (GBS) and a house interior design tool called Autodesk Homestyler. Projects Centaur and Cumulus, currently in beta, use cloud technology to speed processor-intensive simulation.

Portal websites, which also fall under the cloud moniker, are helping facilitate the adoption of high-end tools in what have traditionally been non-technical environments. In his Newport Conference presentation, Stephen Wheat discussed how the Blue Collar Computing (BCC) program offered by the Ohio Supercomputer (OSC) is making advanced modeling tools available to non-HPC people. OSC partner Nimbis Services has created a framework in which the data, software and processing are all available at the click of a mouse. The BCC program from flickrhas partnerships in place with state organizations such as Edison Welding Institute; industry groups such as Polymer Ohio, as well as large manufacturers such as Procter & Gamble.

No more excuses

The cost of entry into the show called digital manufacturing is at an all time low. The top-down HPC technology flow together with the cost-effective accelerative power of GPUs mean that workloads that previously required the power of a supercomputer or high-performance cluster can now be accomplished on multicore (often GPU-CPU hybrid) workstations. And if an in-house system is not in the budget, there's a multitude of cloud computing options to pick from. There truly is a solution for every manufacturer looking to dip its toes into the digital manufacturing waters. But with global competition at an all-time high, not making the plunge could be costly.

Add a Comment

Do NOT follow this link or you will be banned from the site!
Share This