A standard method for testing some of the mechanical properties of materials is to poke them with a sharp point. This “indentation technique” can provide detailed measurements of how the material responds to the point’s force, as a function of its penetration depth.

With advances in nanotechnology during the past two decades, the indentation force can be measured to a resolution on the order of one-billionth of a Newton (a measure of the force approximately equivalent to the force you feel when you hold a medium-sized apple in your hand), and the sharp tip’s penetration depth can be captured to a resolution as small as a nanometer, or about 1/100,000 the diameter of a human hair. Such instrumented nanoindentation tools have provided new opportunities for probing physical properties in a wide variety of materials, including metals and alloys, plastics, ceramics, and semiconductors.

But while indentation techniques, including nanoindentation, work well for measuring some properties, they exhibit large errors when probing plastic properties of materials — the kind of permanent deformation that happens, for example, if you press your thumb into a piece of silly putty and leave a dent, or when you permanently bend a paper clip using your fingers. Such tests can be important in a wide variety of industrial applications, including conventional and digital manufacturing (3-D printing) of metallic structures, material quality assurance of engineering parts, and optimization of performance and cost. However, conventional indentation tests and existing methods to extract critical properties can be highly inaccurate.

Now, an international research team comprising researchers from MIT, Brown University, and Nanyang Technological University (NTU) in Singapore has developed a new analytical technique that can improve the estimation of mechanical properties of metallic materials from instrumented indention, with as much as 20 times greater accuracy than existing methods. Their findings are described today in the Proceedings of the National Academy of Sciences, in a paper combining indentation experiments with computational modeling of materials using the latest machine learning tools.

The team includes co-lead and senior author Ming Dao, a principal research scientist at MIT, and senior author Subra Suresh, MIT Vannevar Bush Professor Emeritus who is president and distinguished university professor at NTU Singapore. Their co-authors are doctoral student Lu Lu and Professor George Em Karniadakis of Brown University and research fellow Punit Kumar and Professor Upadrasta Ramamurty of NTU Singapore.

“Small” challenges beyond elasticity

“Indentation is a very good method for testing mechanical properties,” Dao says, especially in cases where only small samples are available for testing. “When you try to develop new materials, you often have only a small quantity, and you can use indentation or nanoindentation to test really small quantities of materials,” he says.

Such testing can be quite accurate for elastic properties — that is, situations where the material bounces back to its original shape after having been poked. But when the applied force goes beyond the material’s “yield strength” — the point at which the poking leaves a lasting mark on the surface — this is called plastic deformation, and traditional indentation testing becomes much less accurate. “In fact, there’s no widely available method that’s being used” that can produce reliable information in such cases, Dao says.

Indentation can be used to determine hardness, but Dao explains that “hardness is only a combination of a material’s elastic and plastic properties. It’s not a ‘clean’ parameter that can be used directly for design purposes. … But properties at or beyond yield strength, the strength denoting the point at which the material begins to deform irreversibly, are important to access the material’s suitability for engineering applications.”

Technique demands smaller amounts of high-quality data

The new method does not require any changes to experimental equipment or operation, but rather provides a way to work with the data to improve the accuracy of its predictions. By using an advanced neural network machine-learning system, the team found that a carefully planned integration of both real experimental data and computer-generated “synthetic” data of different levels of accuracy (a so-called multifidelity approach to deep learning) can produce the kind of quick and simple yet highly accurate data that industrial applications require for testing materials.

Traditional machine learning approaches require large amounts of high-quality data. However, detailed experiments on actual material samples are time-consuming and expensive to conduct. But the team found that doing the neural network training with lots of low-cost synthetic data and then incorporating a relatively small number of real experimental data points — somewhere between three and 20, as compared with 1,000 or more accurate, albeit high-cost, datasets — can substantially improve the accuracy of the outcome. In addition, they utilize established scaling laws to further reduce the number of training datasets needed in covering the parameter space for all engineering metals and alloys.

What’s more, the authors found that the majority of the time-consuming training process can be done ahead of time, so that for evaluating the actual tests a small number of real experimental results can be added for “calibration” training just when they’re needed, and give highly accurate results.

Applications for digital manufacturing and more

These multifidelity deep-learning approaches have been validated using conventionally manufactured aluminum alloys as well as 3-D-printed titanium alloys.

Professor Javier Llorca, scientific director of IMDEA Materials Institute in Madrid, who was not connected with this research, says, “The new approach takes advantage of novel machine learning strategies to improve the accuracy of the predictions and has a large potential for fast screening of the mechanical properties of components manufactured by 3-D printing. It will allow one to discriminate the differences in the mechanical properties in different regions of the 3-D-printed components, leading to more accurate designs.”

Professor Ares Rosakis at Caltech, who also was not connected with this work, says this approach “results in remarkable computational efficiency and in unprecedented predictive accuracy of the mechanical properties. … Most importantly, it provides a previously unavailable, fresh pair of eyes for ensuring mechanical property uniformity as well as manufacturing reproducibility of 3D-printed components of complex geometry for which classical testing is impossible.”

In principle, the basic process they use could be extended and applied to many other kinds of problems involving machine-learning, Dao says. “This idea, I think, can be generalized to solve other challenging engineering problems.” The use of the real experimental data helps to compensate for the idealized conditions assumed in the synthetic data, where the shape of the indenter tip is perfectly sharp, the motion of the indenter is perfectly smooth, and so on. By using “hybrid” data that includes both the idealized and the real-world situations, “the end result is a drastically reduced error,” he says.

The work was supported by the Army Research Laboratory, the U.S. Department of Energy, and the Nanyang Technical University Distinguished University Professorship.

The post Deep learning for mechanical property evaluation appeared first on Artificial Intelligence.

With every new trend comes a counter-trend. And so despite the current excitement over the wonders of artificial intelligence, one company is betting that human intelligence can still deliver solutions for businesses that AI cannot hope to match.

Article One Partners (AOP) is a crowdsourced network of over 42,000 researchers in 170 countries — 42% of whom have graduate degrees in a variety of science, technology, and engineering specialties. The firm got its start uncovering patent-busting prior art for defendants in high-stakes patent infringement suits, where it quickly earned a reputation for finding invalidating prior art in hidden corners of the globe that Google search could never reach — an unpublished Korean-language PhD dissertation, a rural Norwegian library, even in a New York City pawn shop. Their work often found that a “novel invention” wasn’t so novel after all.

But in recent years, AOP’s sleuths have begun to make a name for themselves as an all-purpose “human search engine” that can help businesses solve challenges that algorithm-based search engines cannot, especially in the development and marketing of innovative new products.

Earlier this year, for example, a small manufacturer based in Europe needed to develop a pipe system that could move highly-volatile and abrasive hydrocarbons like solvents and metal cleaning agents safely over long distances. Hydrocarbons tend to destroy everything they touch — park your car in a puddle of gasoline and your tires will swell and eventually rot. So the company needed to invent a new type of material for the pipe works that would be resistant to organic chemical reactions from the liquid passing through it at varying pressures — and yet still be deformable (i.e., able to swell up to twice its width but then reform to its original shape).

A well-formulated search engine string could certainly point to materials already developed, and research already published. But to find a truly novel yet cost-effective solution, the company felt it needed human insight and expertise in multiple scientific and engineering disciplines. So it retained the British-based innovation consultancy The Moon on a Stick, which in turn called upon AOP for help.

According to The Moon on a Stick’s managing partner Sean Warren, the results were impressive. “AOP’s research crowd came back with 142 possible solutions or compositions that would enable the pipes to withstand the volatile hydrocarbon material and perform as needed,’ Warren noted.  “I was quite surprised by the depth and relevance of the technical approaches they uncovered, some of which the client had never even imagined.”

These included a novel approach using nanotechnology, as well as some little-known new research underway at U.S., European, and Asian universities.

AOP also works with large enterprises, even those with vast internal resources like the telecom giant AT&T and the $30 billion technology giant Philips, the latter of which initially retained AOP to assist with its patent function. But as Brian Hinman, the firm’s retiring chief intellectual property officer, explained, the relationship soon expanded. “We now use AOP to identify manufacturing and distribution channels for certain goods, as well as to explore new trends in particular technology domains.”

One new tech area where Philips was considering expanding its R&D effort was Visible Light Communications (VLC), which uses a band of visual light between 400 and 800 THz to send data such as ads to in-store consumers (or potentially, instant replay video to spectators in a football stadium). Philips deployed AOP experts to start digging for everything they could find — products, companies making products, and new cutting-edge research in VLC — that would help the company make a business decision on whether, and how, to invest in VLC or not.

This is where the distinction between algorithms and human judgment becomes crucial. A search engine query can quickly tell you a lot about VLC, its history, a few of the major players, and some published research in the field. But to make a business decision about whether to invest tens of millions of dollars in developing and marketing VLC products, Philips needed the experi8ence, insight, and business judgment of human experts who could assess the size and scope of the market opportunity as well as the best “white space” innovation areas for the firm.

Bet-the-company decisions like that should not be left to an algorithm, said Philips’s Hinman. “AOP produced actionable intelligence that enabled us to make informed decisions regarding innovation focus, invention generation, and potential acquisitions.”

To be sure, the robust AI systems now being designed and implemented do more than simply answer search queries. They can also manage systems, conduct operations, and take action. But fundamentally — at least so far — these are differences mostly of degree, not kind.

In any event, for challenges that quite literally require boots on the ground, even the most advanced AI system won’t be able to compete with a network of human sleuths. AOP’s CEO Peter Vanderheyden offered one example:

“We were engaged by a global licensing organization for one of the world’s biggest consumer products,” he recalled. “They asked us to find out where unlicensed devices were being sold around the world. Now, Google could point to all kinds of articles about counterfeit products in China or in India. It could also give you estimates of the losses due to counterfeit product sales. But that only tells this licensing organization what they already know, right?

“So we asked our researchers to go out and actually knock on doors,” he continued. “We had them go into their local stores, in whatever country they were located, and take six pictures of every box containing a device that featured this advertised consumer technology — one photo of each side of the box. The goal was to see if the package displayed the proper license label.”

To no one’s surprise, AOP sleuths produced photos of quite a few unlicensed products around the world. “And mind you, this was unbiased, third party, time-stamped evidence,” he added. “Very admissible in court. Which you better believe this licensing organization made sure to mention when it contacted those unlicensed vendors.”

Vanderheyden claimed that AOP’s work helped the licensing organization collect millions of dollars in new licensing revenues, though he declined to be more specific. “We also helped identify ways to improve their licensing control practices to reduce problems,” he added.

AOP’s latest bet on human intelligence was the launch last month of a new TalentSource service, offering qualified expert technologists from its crowd on a contract basis to companies. The aim here is to fill a growing need within companies for expertise in new or adjacent technologies outside their core R&D competence that industry convergence is increasingly forcing them to contend with. TalentSource enables these firms to bring in the talent needed to explore these new technological areas without having to invest yet in hiring full-time staff.

What’s unique about TalentSource compared to traditional technology consulting firms in the industry? AOP’s Vanderheyden claims it’s the depth of its bench of subject matter experts — again, 42,000 experts, almost half of whom have advanced degrees — as well as the flexible on-demand nature of their availability.

Whatever happens with Article One Partners and its various ventures in HI (human intelligence), it does seem clear that despite the enormous promise of AI,  there will always be some challenges that require human judgment, expertise, and insight to deal with effectively.

The post Where human intelligence outperforms AI appeared first on Artificial Intelligence.