More stories

  • in

    New shape memory alloy discovered through artificial intelligence framework

    Funded by the National Science Foundation’s Designing Materials to Revolutionize Our Engineering Future (DMREF) Program, researchers from the Department of Materials Science and Engineering at Texas A&M University used an Artificial Intelligence Materials Selection framework (AIMS) to discover a new shape memory alloy. The shape memory alloy showed the highest efficiency during operation achieved thus far for nickel-titanium-based materials. In addition, their data-driven framework offers proof of concept for future materials development.
    Shape memory alloys are utilized in various fields where compact, lightweight and solid-state actuations are needed, replacing hydraulic or pneumatic actuators because they can deform when cold and then return to their original shape when heated. This unique property is critical for applications, such as airplane wings, jet engines and automotive components, that must withstand repeated, recoverable large-shape changes.
    There have been many advancements in shape memory alloys since their beginnings in the mid-1960s, but at a cost. Understanding and discovering new shape memory alloys has required extensive research through experimentation and ad-hoc trial and error. Despite many of which have been documented to help further shape memory alloy applications, new alloy discoveries have occurred in a decadal fashion. About every 10 years, a significant shape memory alloy composition or system has been discovered. Moreover, even with advances in shape memory alloys, they are hindered by their low energy efficiency caused by incompatibilities in their microstructure during the large shape change. Further, they are notoriously difficult to design from scratch.
    To address these shortcomings, Texas A&M researchers have combined experimental data to create an AIMS computational framework capable of determining optimal materials compositions and processing these materials, which led to the discovery of a new shape memory alloy composition.
    “When designing materials, sometimes you have multiple objectives or constraints that conflict, which is very difficult to work around,” said Dr. Ibrahim Karaman, Chevron Professor I and materials science and engineering department head. “Using our machine-learning framework, we can use experimental data to find hidden correlations between different materials’ features to see if we can design new materials.”
    The shape memory alloy found during the study using AIMS was predicted and proven to achieve the narrowest hysteresis ever recorded. In other words, the material showed the lowest energy loss when converting thermal energy to mechanical work. The material showcased high efficiency when subject to thermal cycling due to its extremely small transformation temperature window. The material also exhibited excellent cyclic stability under repeated actuation.
    A nickel-titanium-copper composition is typical for shape memory alloys. Nickel-titanium-copper alloys typically have titanium equal to 50% and form a single-phase material. Using machine learning, the researchers predicted a different composition with titanium equal to 47% and copper equal to 21%. While this composition is in the two-phase region and forms particles, they help enhance the material’s properties, explained William Trehern, doctoral student and graduate research assistant in the materials science and engineering department and the publication’s first author.
    In particular, this high-efficiency shape memory alloy lends itself to thermal energy harvesting, which requires materials that can capture waste energy produced by machines and put it to use, and thermal energy storage, which is used for cooling electronic devices.
    More notably, the AIMS framework offers the opportunity to use machine-learning techniques in materials science. The researchers see potential to discover more shape memory alloy chemistries with desired characteristics for various other applications.
    “It is a revelation to use machine learning to find connections that our brain or known physical principles may not be able to explain,” said Karaman. “We can use data science and machine learning to accelerate the rate of materials discovery. I also believe that we can potentially discover new physics or mechanisms behind materials behavior that we did not know before if we pay attention to the connections machine learning can find.”
    Other contributors include Dr. Raymundo Arróyave and Dr. Kadri Can Atli, professors in the materials science and engineering department, and materials science and engineering undergraduate student Risheil Ortiz-Ayala.
    “While machine learning is now widely used in materials science, most approaches to date focus on predicting the properties of a material without necessarily explaining how to process it to achieve target properties,” said Arróyave. “Here, the framework looked not only at the chemical composition of candidate materials, but also the processing necessary to attain the properties of interest.”
    Story Source:
    Materials provided by Texas A&M University. Original written by Michelle Revels. Note: Content may be edited for style and length. More

  • in

    Newly proposed search strategies improve computational cost of the bicycle-sharing problem

    Bicycle sharing systems (BSSs) are transport solutions wherein users can rent a bicycle from a depot or ‘port,’ travel, and then return the bike to the same port or different port. BSSs are growing in popularity around the world because they are eco-friendly, reduce traffic congestion, and offer added health benefits to users. But eventually, a port becomes either full or empty in a BSS. This means that users are no longer able to rent a bike (when empty) or return one (when full). To address this issue, bikes need to be rebalanced among the ports in a BSS so that users are always able to use them. This rebalancing must also be carried out in a way that is beneficial to BSS companies so that they can reduce labor costs, as well as carbon emissions from rebalancing vehicles.
    There are several existing approaches to BSS rebalancing, however, most solution algorithms are computationally expensive and take a lot of time to find an ‘exact’ solution in cases where there are a large number of ports. Even finding an approximate solution is computationally expensive. Previously, a research team led by Prof. Tohru Ikeguchi from Tokyo University of Science proposed a ‘multiple-vehicle bike sharing system routing problem with soft constraints’ (mBSSRP-S) that can find the shortest travel times for multiple bike rebalancing vehicles with the caveat that the optimal solution can sometimes violate the real-world limitations of the problem. Now, in a recent study published in MDPI’s Applied Sciences, the team has proposed two strategies to search for approximate solutions to the mBSSRP-S that can reduce computational costs without affecting performance. The research team also featured PhD student Ms. Honami Tsushima of Tokyo University of Science and Prof. Takafumi Matsuura of Nippon Institute of Technology.
    Describing their research, Prof. Ikeguchi says, “Earlier, we had proposed the mBSSRP-S and that offered improved performance as compared to our original mBSSRP, which did not allow the violation of constraints. But the mBSSRP-S also increased the overall computational cost of the problem because it had to calculate both the feasible and infeasible solutions of the mBSSRP. Therefore, we have now proposed two consecutive search strategies to address this problem.”
    The proposed search strategies look for feasible solutions in a much shorter period of time as compared to the one originally proposed with mBSSRP-S. The first strategy focuses on reducing the number of ‘neighboring’ solutions (solutions that are numerically close to a solution to the optimization problem) before finding a feasible solution. The strategy employs two well-known algorithms called ‘Or-opt’ and ‘CROSS-exchange,’ to reduce the overall time taken to compute a solution. The feasible solution here refers to values that satisfy the constraints of mBSSRP.
    The second strategy changes the problem to be solved based on the feasible solution to either the mBSSRP problem or the mBSSRP-S problem and then searches for good near-optimal solutions in a short time by either Or-opt or CROSS-exchange.
    The research team then performed numerical experiments to evaluate the computational cost and performance of their algorithms. “With the application of these two strategies, we have succeeded in reducing computational time while maintaining performance,” reveals Prof. Ikeguchi. “We also found that once we calculated the feasible solution, we could find short travel times for the rebalancing vehicles quickly by solving the hard constraint problem, mBSSRP, instead of mBSSRP-S.”
    The popularity of BSSs is only expected to grow in the future. The new solution-search strategies proposed here will go a long way towards realizing convenient and comfortable BSSs that benefit users, companies, and the environment.
    Story Source:
    Materials provided by Tokyo University of Science. Note: Content may be edited for style and length. More

  • in

    Researchers now able to predict battery lifetimes with machine learning

    Technique could reduce costs of battery development.
    Imagine a psychic telling your parents, on the day you were born, how long you would live. A similar experience is possible for battery chemists who are using new computational models to calculate battery lifetimes based on as little as a single cycle of experimental data.
    In a new study, researchers at the U.S. Department of Energy’s (DOE) Argonne National Laboratory have turned to the power of machine learning to predict the lifetimes of a wide range of different battery chemistries. By using experimental data gathered at Argonne from a set of 300 batteries representing six different battery chemistries, the scientists can accurately determine just how long different batteries will continue to cycle.
    In a machine learning algorithm, scientists train a computer program to make inferences on an initial set of data, and then take what it has learned from that training to make decisions on another set of data.
    “For every different kind of battery application, from cell phones to electric vehicles to grid storage, battery lifetime is of fundamental importance for every consumer,” said Argonne computational scientist Noah Paulson, an author of the study. “Having to cycle a battery thousands of times until it fails can take years; our method creates a kind of computational test kitchen where we can quickly establish how different batteries are going to perform.”
    “Right now, the only way to evaluate how the capacity in a battery fades is to actually cycle the battery,” added Argonne electrochemist Susan “Sue” Babinec, another author of the study. “It’s very expensive and it takes a long time.”
    According to Paulson, the process of establishing a battery lifetime can be tricky. “The reality is that batteries don’t last forever, and how long they last depends on the way that we use them, as well as their design and their chemistry,” he said. “Until now, there’s really not been a great way to know how long a battery is going to last. People are going to want to know how long they have until they have to spend money on a new battery.” More

  • in

    'Nanomagnetic' computing can provide low-energy AI

    Researchers have shown it is possible to perform artificial intelligence using tiny nanomagnets that interact like neurons in the brain.
    The new method, developed by a team led by Imperial College London researchers, could slash the energy cost of artificial intelligence (AI), which is currently doubling globally every 3.5 months.
    In a paper published today in Nature Nanotechnology, the international team have produced the first proof that networks of nanomagnets can be used to perform AI-like processing. The researchers showed nanomagnets can be used for ‘time-series prediction’ tasks, such as predicting and regulating insulin levels in diabetic patients.
    Artificial intelligence that uses ‘neural networks’ aims to replicate the way parts of the brain work, where neurons talk to each other to process and retain information. A lot of the maths used to power neural networks was originally invented by physicists to describe the way magnets interact, but at the time it was too difficult to use magnets directly as researchers didn’t know how to put data in and get information out.
    Instead, software run on traditional silicon-based computers was used to simulate the magnet interactions, in turn simulating the brain. Now, the team have been able to use the magnets themselves to process and store data — cutting out the middleman of the software simulation and potentially offering enormous energy savings.
    Nanomagnetic states
    Nanomagnets can come in various ‘states’, depending on their direction. Applying a magnetic field to a network of nanomagnets changes the state of the magnets based on the properties of the input field, but also on the states of surrounding magnets. More

  • in

    Bye, bye, biopsy? Handheld device could painlessly identify skin cancers

    Skin biopsies are no fun: doctors carve away small lumps of tissue for laboratory testing, leaving patients with painful wounds that can take weeks to heal. That’s a price worth paying if it enables early cancer treatment. However, in recent years, aggressive diagnostic efforts have seen the number of biopsies grow around four times faster than the number of cancers detected, with about 30 benign lesions now biopsied for every case of skin cancer that’s found.
    Researchers at Stevens Institute of Technology are now developing a low-cost handheld device that could cut the rate of unnecessary biopsies in half and give dermatologists and other frontline physicians easy access to laboratory-grade cancer diagnostics. “We aren’t trying to get rid of biopsies,” said Negar Tavassolian, director of the Bio-Electromagnetics Laboratory at Stevens. “But we do want to give doctors additional tools and help them to make better decisions.”
    The team’s device uses millimeter-wave imaging — the same technology used in airport security scanners — to scan a patient’s skin. (In earlier work, Tavassolian and her team had to work with already biopsied skin for the device to detect if it was cancerous.)
    Healthy tissue reflects millimeter-wave rays differently than cancerous tissue, so it’s theoretically possible to spot cancers by monitoring contrasts in the rays reflected back from the skin. To bring that approach into clinical practice, the researchers used algorithms to fuse signals captured by multiple different antennas into a single ultrahigh-bandwidth image, reducing noise and quickly capturing high-resolution images of even the tiniest mole or blemish.
    Spearheaded by Amir Mirbeik Ph.D. ’18, the team used a tabletop version of their technology to examine 71 patients during real-world clinical visits, and found their methods could accurately distinguish benign and malignant lesions in just a few seconds. Using their device, Tavassolian and Mirbeik could identify cancerous tissue with 97% sensitivity and 98% specificity — a rate competitive with even the best hospital-grade diagnostic tools.
    “There are other advanced imaging technologies that can detect skin cancers, but they’re big, expensive machines that aren’t available in the clinic,” said Tavassolian, whose work appears in the March 23 issue of Scientific Reports. “We’re creating a low-cost device that’s as small and as easy to use as a cellphone, so we can bring advanced diagnostics within reach for everyone.”
    Because the team’s technology delivers results in seconds, it could one day be used instead of a magnifying dermatoscope in routine checkups, giving extremely accurate results almost instantly. “That means doctors can integrate accurate diagnostics into routine checkups, and ultimately treat more patients,” said Tavassolian.
    Unlike many other imaging methods, millimeter-wave rays harmlessly penetrate about 2mm into human skin, so the team’s imaging technology provides a clear 3D map of scanned lesions. Future improvements to the algorithm powering the device could significantly improve mapping of lesion margins, enabling more precise and less invasive biopsying for malignant lesions.
    The next step is to pack the team’s diagnostic kit onto an integrated circuit, a step that could soon allow functional handheld millimeter-wave diagnostic devices to be produced for as little as $100 a piece — a fraction of the cost of existing hospital-grade diagnostic equipment. The team is already working to commercialize their technology and hopes to start putting their devices in clinicians’ hands within the next two years.
    “The path forward is clear, and we know what we need to do,” said Tavassolian. “After this proof of concept, we need to miniaturize our technology, bring the price down, and bring it to the market.”
    Story Source:
    Materials provided by Stevens Institute of Technology. Note: Content may be edited for style and length. More

  • in

    The quest for an ideal quantum bit

    New qubit platform could transform quantum information science and technology.
    You are no doubt viewing this article on a digital device whose basic unit of information is the bit, either 0 or 1. Scientists worldwide are racing to develop a new kind of computer based on use of quantum bits, or qubits.
    In a recent Nature paper, a team led by the U.S. Department of Energy’s (DOE) Argonne National Laboratory has announced the creation of a new qubit platform formed by freezing neon gas into a solid at very low temperatures, spraying electrons from a light bulb’s filament onto the solid, and trapping a single electron there. This system shows great promise to be developed into ideal building blocks for future quantum computers.
    To realize a useful quantum computer, the quality requirements for the qubits are extremely demanding. While there are various forms of qubits today, none of them is ideal.
    What would make an ideal qubit? It has at least three sterling qualities, according to Dafei Jin, an Argonne scientist and the principal investigator of the project.
    It can remain in a simultaneous 0 and 1 state (remember the cat!) over a long time. Scientists call this long “coherence.” Ideally, that time would be around a second, a time step that we can perceive on a home clock in our daily life. More

  • in

    Deep learning model to predict adverse drug-drug interactions

    Prescriptions for multiple drugs, or polypharmacy, is often recommended for the treatment of complex diseases. However, upon ingestion, multiple drugs may interact in an undesirable manner, resulting in severe adverse effects or decreased clinical efficacy. Early detection of such drug-drug interactions (DDIs) is therefore essential to prevent patients from experiencing adverse effects.
    Currently, computational models and neural network-based algorithms examine prior records of known drug interactions and identify the structures and side effects they are associated with. These approaches assume that similar drugs have similar interactions and identify drug combinations associated with similar adverse effects.
    Although understanding the mechanisms of DDIs at a molecular level is essential to predict their undesirable effects, current models rely on structures and properties of drugs, with predictive range limited to previously observed interactions. They do not consider the effect of DDIs on genes and cell functionality.
    To address these limitations, Associate Professor Hojung Nam and Ph.D. candidate Eunyoung Kim from the Gwangju Institute of Science and Technology in South Korea developed a deep learning-based model to predict DDIs based on drug-induced gene expression signatures. These findings were published in the Journal of Cheminformatics on March 4, 2022.
    The DeSIDE-DDI model consists of two parts: a feature generation model and a DDI prediction model. The feature generation model predicts a drug’s effect on gene expression by considering both the structure and properties of the drug while the DDI prediction model predicts various side effects resulting from drug combinations.
    To explain the key features of this model, Prof. Nam explains, “Our model considers the effects of drugs on genes by utilizing gene expression data, providing an explanation for why a certain pair of drugs cause DDIs. It can predict DDIs for currently approved drugs as well as for novel compounds. This way, the threats of polypharmacy can be resolved before new drugs are made available to the public.”
    What’s more, since all compounds do not have drug-treated gene expression signatures, this model uses a pre-trained compound generation model to generate expected drug-treated gene expressions.
    Discussing its real-life applications, Prof. Nam remarks, “This model can discern potentially dangerous drug pairs, acting as a drug safety monitoring system. It can help researchers define the correct usage of the drug in the drug development phase.”
    A model with such potential will truly revolutionize how the safety of novel drugs is established in the future.
    Story Source:
    Materials provided by GIST (Gwangju Institute of Science and Technology). Note: Content may be edited for style and length. More

  • in

    Taste of the future: Robot chef learns to 'taste as you go'

    A robot ‘chef’ has been trained to taste food at different stages of the chewing process to assess whether it’s sufficiently seasoned.
    Working in collaboration with domestic appliances manufacturer Beko, researchers from the University of Cambridge trained their robot chef to assess the saltiness of a dish at different stages of the chewing process, imitating a similar process in humans.
    Their results could be useful in the development of automated or semi-automated food preparation by helping robots to learn what tastes good and what doesn’t, making them better cooks.
    When we chew our food, we notice a change in texture and taste. For example, biting into a fresh tomato at the height of summer will release juices, and as we chew, releasing both saliva and digestive enzymes, our perception of the tomato’s flavour will change.
    The robot chef, which has already been trained to make omelettes based on human taster’s feedback, tasted nine different variations of a simple dish of scrambled eggs and tomatoes at three different stages of the chewing process, and produced ‘taste maps’ of the different dishes.
    The researchers found that this ‘taste as you go’ approach significantly improved the robot’s ability to quickly and accurately assess the saltiness of the dish over other electronic tasting technologies, which only test a single homogenised sample. The results are reported in the journal Frontiers in Robotics & AI. More