Computer vision researchers use machine learning to train computers in visually recognizing objects – but very few apply machine learning to mechanical parts such as gearboxes, bearings, brakes, clutches, motors, nuts, bolts and washers.
A team of Purdue University mechanical engineers has created the first comprehensive open-source annotated database of more than 58,000 3D mechanical parts, designed to help researchers apply machine learning to those parts in actual machines.
“We are in the deep learning era, using computers to search for things visually,” said Karthik Ramani, Purdue’s Donald W. Feddersen Distinguished Professor of Mechanical Engineering. “But no one is focusing on the parts that go into machines: pipes, bearings, motors, washers, nuts and bolts, etc. Those are the things that are important to us as engineers and manufacturers. We want to be able to point a camera at a real-world part, and have the computer tell us everything about that part or design.”
Ramani’s team experimented with visual search for parts in the early 2000s, but computing power and machine learning techniques were not yet sufficiently advanced. In the years since, researchers have learned that building a solid dataset is all about both quality and quantity.
“Deep learning is data hungry,” Ramani said. “It needs a lot of examples for the computer to learn what humans mean and how things relate to each other. That means we needed a lot of 3D models of parts which also required an underlying engineering classification.”
The team began by partnering with a French company called TraceParts, which gave Purdue researchers access to their database of 3D engineering parts. The team collaborated with the University of Texas, Austin assistant professor Qixing Huang to scour other databases for similar 3D models. They ended up compiling a database of 58,696 mechanical components. A video about the database is available on YouTube.
But a database is no good without good data. Ramani’s team organized the parts by establishing a hierarchical taxonomy of 68 classes, based on the International Classification for Standards, a system of technical standards created and maintained by the International Organization for Standardization.
“Now when a computer sees a picture of a seal component, it will know that it fits in the category of dynamic seals and then, more specifically, under composite seals,” Ramani said.
The researchers now have published their open-source database, inviting computer vision and machine learning researchers to access it and create their own experiments. They showcased their work at the 16th European Conference on Computer Vision in August.
With a benchmark dataset now established, what is the future of machine learning for making machines?
“We see many real-world situations for this technology,” Ramani said. “Imagine you’re working maintenance in a factory, and you’re replacing a part of a machine. You can point a camera at the part, and the computer will recognize it, and instantly give you all the specifications of that part – what it’s called, what it connects to, and where they are physically stored in the factory. This could even happen through augmented reality glasses; you could have your company’s entire visual catalog instantly at your fingertips and learn how to fix things or order parts.
“There are a lot of challenges in machine learning. But you can’t deal with the challenge unless you know how to test how well you’re doing. We are proud to have created the first large-scale benchmark dataset of annotated 3D mechanical components in the world, and we hope to advance visual search in the deep learning era.”
This research is being conducted at the Convergence Design Laboratory at Purdue University. The work is partially supported by the National Science Foundation.
Electric vehicles (EVs) will replace light-duty internal-combustion-engine vehicles (ICEVs) that run on gasoline and diesel in the coming years. The shift is necessary to combat climate change. It could also upend global competition in the world’s largest manufacturing sector. Most leading vehicle-producing regions and nations perceive the opportunity—and the threat—this paradigm shift represents. They are taking action in response—with one big exception: the United States. It’s high time that the federal government adopted a serious EV manufacturing strategy.
Climate change is here. Wildfires and severe storms exacerbated by climate change have ravaged diverse regions of the United States in recent months. To prevent such effects from getting much worse, the world needs to rapidly reduce and ultimately eliminate carbon dioxide emissions. ICEVs cause about one-sixth of U.S. emissions and one-tenth globally.
These facts have prompted many governments to announce that sales of ICEVs will be phased out. Norway is the first-mover, eliminating such sales after 2025. The United Kingdom recently moved forward its phase-out to 2035, the same year targeted by California. Sales of ICEVs in many other European countries will end in the 2030s, with France aiming for 2040. China plans to have “new energy vehicles” like EVs comprise at least 50% of its car market by 2035, with the remainder restricted to hybrids, but its vow a few weeks ago to achieve carbon neutrality by 2060 will probably accelerate that timetable.
High-minded promises about the distant future are often mere virtue-signaling by today’s office-holders, conveniently forgotten by their successors. But in this case, the pledges may well be redeemed. The reason is that EVs are already reasonably good substitutes for ICEVs from the consumer perspective, and they are improving rapidly. Key barriers to widespread adoption, including initial cost, range, charging time, and charging infrastructure, are likely to be overcome in the next few years.
In the shadow of government action around the world, the market is starting to drive the EV transition. Tesla has shown what is possible. Over the summer, its soaring market capitalization surpassed Detroit’s Big Three, Daimler, and Honda—combined. Other EV start-ups like Fisker and Lordstown are seeking to follow in Tesla’s footsteps and receiving outlandish valuations as well. Established automakers have taken notice. Volkswagen, compelled by its “Dieselgate” scandal, is pumping 60 billion euros into EVs. GM CEO Mary Barra promises to “offer EVs from every brand, every segment, every body style, every price point and in every part of the country,” even though EVs today make up only about 2% of U.S. auto sales.
Whether the insurgents or the incumbents will ultimately gain the upper hand in the revamped industry, and which members of each group will survive, are matters of intense conjecture among analysts. EVs require different components, suppliers, skills, and business models than today’s cars, challenging the likes of GM and VW, but they also require coordination and integration on a massive scale, stretching the managerial aptitude of the newcomers.
But this story isn’t just about companies. It’s about places and people, too. Vehicle assembly plants are big employers, linchpins of local economies. So are plants that make major components, whether EV batteries and motors or ICEV engines, fuel systems, and transmissions. Lower tiers of the auto supply chain provide livelihoods for myriad small- and medium-sized manufacturers, which often cluster near big plants.
As the production paradigm shifts in the coming years, these regional clusters—whether in Michigan or Tennessee in the United States, or in Aichi prefecture in Japan or Baden-Wurttemberg state in Germany—could face a shake-out as severe as that facing the companies. Such clusters are concentrations of specialized facilities, skills, and relationships that may be difficult to adapt to EV production. The kind of shock that devastated Detroit in the 1970s when fuel-efficient imports first made inroads into the U.S. market could be replicated in many places around the world in the 2020s.
As the history of Detroit illustrates, the vitality of auto-dependent regions weighs heavily on national economies. ICEV manufacturing is one of the largest, if not the largest, manufacturing industries in the European Union, the United Kingdom, Japan, Korea, China, and Brazil as well the United States.
Motor vehicles and automotive parts also represent about 10 percent of global trade. As EVs and their components loom larger in the industry, its trade patterns could shift, altering which countries benefit most. Foreign direct investment, too, could reorient the location of value capture in this massive industry, as it did when “voluntary” export restraints drew foreign manufacturers to the United States in the 1980s.
By almost any definition, the auto industry is strategically important to the world’s leading economies. The United States’ willingness to bail out GM and Chrysler during the 2008 financial crisis illustrates this fact. National governments must pay attention to the EV transition or they will be putting their economies—and by extension, their political viability—at risk.
China is certainly paying attention. Beijing sees the transition as an opportunity not only to establish a stronger presence for domestic firms in its home market, but also as a pathway to export success. The fact that no Chinese auto firms have achieved global scale, even with the benefit of technology transferred under duress by foreign partners, is a sore point for the central government. To seize the moment, it has designated “new energy vehicles” a priority for economic development, prompting massive investments by state-owned institutions as well as private investors. Chinese companies now dominate many parts of the EV supply chain and some have begun to break through globally. CATL, a 10-year-old battery giant based in Fujian province, for instance, is building a factory in Germany to supply BMW, Mercedes, and European manufacturers.
European governments, in turn, view China’s foothold in the EV sector, as well as Tesla’s meteoric rise, with concern. The European Union accounted for more than half of global vehicle and parts exports, so it has the most to lose from shifts in global auto trade. In 2018, the European Commission adopted a comprehensive policy framework to support the establishment of a complete EV battery value chain in Europe. The Commission’s investment arm, the European Investment Bank, along with EU member-states France and Germany, has invested in Northvolt, a new company that is building battery “giga-factories” in Sweden and Germany.
Japan and South Korea have also adopted EV manufacturing strategies. Although these nations are home to companies like Panasonic and Samsung, which are among the world’s leading battery manufacturers, many of their car companies are having to shift their strategies. For example, Toyota, the pioneer of hybrid vehicles, had long been committed to hydrogen fuel-cells for EV propulsion. It’s now clear that batteries will become the standard instead, leaving Toyota and many other Japanese and Korean companies scrambling to catch up.
Of all the major auto powers, only the United States lacks a serious EV manufacturing strategy. It had the beginnings of one in 2009. The Obama administration provided low-interest loans to build Tesla’s assembly plant in Fremont, California, and renovate Nissan’s Leaf EV plant in Smyrna, Tennessee. Battery and EV component manufacturing received some $2 billion cost-shared grants under the 2009 Recovery Act, which also included a 30 percent tax credit for investment in advanced energy manufacturing. But further extension of these efforts was blocked by Congress after 2010, and the Trump administration showed little interest in reviving them (although its Energy Storage Grand Challenge offered a glimmer of hope in 2020).
Continuing on without a strategy and leaving the location of EV production to “market forces” would simply lead to greater reliance on imports. Strategic, and sometimes unfair and illegal, behavior by other countries would ensure that U.S. EV production is weakened. Even when economic logic strongly favored production near the point of sale, Chinese, European, and other subsidies would often override it.
In the absence of federal action, the states have stepped in, but that is a poor substitute for a national strategy. States have long competed with one another to lure auto factories, a pattern that has continued into the EV era. Tesla, for instance, received a package worth an estimated $1.25 billion to build its first giga-factory in Nevada in 2014. Interstate competition frequently leads to incentive packages for producers that are far in excess of what international competition actually requires, while undermining the ability of states and localities to deliver vital services to citizens. A national strategy would limit such waste.
An even bigger problem with leaving EV manufacturing strategy to the states is their inability to drive innovation. Successful industrial clusters extend far beyond manufacturing plants and encompass R&D, supply chain integration, and workforce development activities that allow plants to outperform their rivals elsewhere. Even large states like California and New York, both of which are pursuing EV manufacturing strategies, lack the capacity to effectively implement policy in all of these areas. Federal laboratories, programs, and financial resources are crucial to fill the gaps.
A national strategy is also needed to ensure that EV manufacturing is coordinated with infrastructure innovations. Vehicle-to-grid integration could make the electricity system more efficient and reduce emissions. Adoption of EVs by government and corporate fleets, including for ride-sharing and public transit, could do the same. Recycling may hold the key to avoiding dependence on critical minerals as well as limiting lifecycle environmental impacts. The success of all three may depend on the adoption of national standards by manufacturers as well as public infrastructure investment.
A serious national EV manufacturing strategy would require money. President-elect Biden, who promised to adopt such a strategy during his 2020 campaign, could seek to draw on the $17.7 billion in loan authority available under the federal Advanced Technology Vehicle Manufacturing (ATVM) program or ask Congress to revive the Advanced Energy Manufacturing Tax Credit to support investment. The new administration could create new public-private manufacturing innovation institutes within the Manufacturing USA network, expand federal programs that accelerate the uptake of innovations like the Advanced Research Projects Agency—Energy’s (ARPA-E)’s new SCALEUP program, and invest in training and apprenticeship programs so that auto workers are prepared for the transition.
A serious national EV manufacturing strategy would need a trade component to fend off unfair foreign practices while ensuring U.S. products got a fair shot at markets abroad. The United States might make common cause with allies in this respect to counter predatory Chinese behavior, including by banning Chinese imports based on unfair practices or subsidies Federal procurement authority could be used to be ensure a level playing field for American-made EVs as well.
Federal procurement is also a key tool to drive rapid growth in domestic demand for EVs. State mandates, fuel economy regulations, tax incentives, and carbon pricing might be used to this end, too. It’s certainly true that without robust domestic demand, any EV manufacturing strategy is bound to fail. But it’s also true that relying on demand-pull policies alone is likely to be a losing strategy for domestic producers.
The solar panel industry is a cautionary tale in this regard. Federal- and state-incentivized demand has led to soaring installation over the past two decades, but virtually all of it has been supplied by imports from China and other Asian locations. China now dominates every part of the solar PV supply chain, gaining massive scale and learning advantages. To avoid making the same outcome in the EV industry, federal policy must address both supply and demand.
The starting gun for the race to be a part of the auto industry’s future has sounded. The United States is already lagging behind. Unless the federal government gets serious, the nation may be stuck there for good.
David M. Hart is a senior fellow at the Information Technology and Innovation Foundation (ITIF) and professor of public policy and director of the Center for Science, Technology and Innovation Policy at George Mason University.
This article originally appeared on the ITIF’s website. It is used with permission.
Source: https://www.industryweek.com/technology-and-iiot/emerging-technologies/article/21147734/machine-learning-for-machine-parts
