Regeneron International Science and Engineering Fair (ISEF) 2021 was held virtually from May 16 to May 21, bringing together over 1800 finalists from nearly 400 affiliate fairs in 64 countries. Engineering.com had the opportunity to chat with two big winners: Brian Minnick, who won first place in Engineering Mechanics for creating a self-replicating 3D printer; and John Benedict Estrada, who received the $50,000 Gordon E. Moore Award for his AI model that predicted drought stress in plants.
A Self-Replicating 3D Printer with Earth and Space Applications
Brian Minnick wanted to change the notion that self-replicating machines—or universal constructors—are an element only of science fiction. With the open-source RepRap project as his starting point, Minnick was able to achieve 100 percent 3D printability with his machine; i.e., every part of his 3D printer could be manufactured by another 3D printer of the same type. He also made significant headway towards self-assembly, with the machine’s self-manufacturing capabilities making it the first assisted replicator to date.
“The novel conductive material was the part of the project that took me the longest to solve,” said Minnick. “It was also the most important piece of the whole project, since each of the four project goals [generating mechanical power, data storage and interpretation, and the hot end] required electronics.”
Minnick began by testing a commercially available conductive material, whose conductivity proved to be poor. He tried developing a novel composite conductive material by adding fine wire bits into the commercially available material, but this only improved conductivity by 50 percent. His next attempts involved innately conducting polymers and printing with solder wire directly, but these experiments were not successful. Minnick then thought of making his solder globules smaller when chaining them together to form a solid conductor. This led to solder paste, which could be sintered into a solid trace using a hot air gun or through controlled heating in an oven. The resulting solution demonstrated 98.3 percent less resistivity than the best commercial alternative material.
For Earth-based applications of his 3D printer, Minnick designed a 3D printed wind turbine for power generation (to be used in environments with atmosphere).
Then there was the task of programming the printer for parts.
“To print a model, the printer must have a data strip representing the model to be printed,” detailed Minnick. “The data strip has columns and rows, where each row corresponds to a linear movement of the printer, and each column corresponds to one particular function of the data strip. Within each row, each column can be toggled on or off, and the distribution of “activated” columns encodes the rotation speed and direction for all of the motors on the printer. This acts as a heading—the direction the printer will move in. The length of each row encodes how long the printer moves in that direction.
“I also created a data strip generator program, which takes each linear movement in a digital model and encodes it in a row in an output data strip. This allows any digital model to be transformed into a representative data strip to be printed on the fully 3D printed 3D printer. The beauty of this control system is that it can be used with any kinematics; it doesn’t have to be a 3D printer. The same motor controller and data strip system can be used to control robotic arms, specialized data strip duplicators, or a method of loading and unloading data strips to print different parts. This modularity was a key goal of my design process.”
When used as a self-replicating spacecraft—Minnick’s ultimate dream for his 3D printer—the machine must have the ability to generate and process the material itself. Use of a biopolymer to build 95 percent of the 3D printer makes it possible for materials to be grown on other planets. As for the other five percent, the printer uses small amounts of a PEEK polymer, a polarizable magnetic material made from iron particles and plastic, and a conductive solder paste material made from an alloy of tin and other metals.
Minnick eliminated complex assembly processes by designing the kinematics of the 3D printer to be print-in-place.
“Using a conventional 3D printer I designed and built, about 80 percent of the machine can be 3D printed in less than a day, and for less than $5 worth of materials,” said Minnick. “Since the 3D printer cannot make items larger than itself in one piece, there must be some assembly done, but this is minimized by the print-in-place design.”
According to Minnick, his self-replicating 3D printer has a wide range of applications—both on Earth and in space.
On Earth, his machines can exponentially increase manufacturing capacity, quickly responding to spikes in demand of critically needed equipment.
“Since they can also build their own power and other infrastructure, they can bring the benefits of modern manufacturing technology to remote regions, raising the standard of living,” said Minnick. “Because the printer is made from a biopolymer, it is more biodegradable than petroleum-based polymers, and thus will not contribute toward plastic pollution in comparison with alternatives.”
Minnick’s dreams for space applications are even more ambitious.
“Self-replicating space probes can explore the galaxy in record time, allowing us insight into the processes that created, continued, and will eventually consume our planet—potentially helping us battle issues like global warming,” asserted Minnick. “These newly discovered worlds can reveal new biology, physics, and potentially places for future human expansion. The ability of the self-replicating spacecraft to exponentially expand across the galaxy allows them to achieve these goals in a fraction of the time compared to traditional methods. Finally, self-replicating lunar factories can process and refine materials for Earth, bringing highly polluting industries off of Earth while providing more raw materials. These factories can also mass produce satellites and spacecraft at virtually no cost to Earth. Solar power satellites created in this way can provide almost limitless, clean energy to Earth, resolving our dependence on fossil fuels.”
So, how did Minnick put the project together during the pandemic?
“I worked solely out of my bedroom,” revealed Minnick. “The nature of the project demanded that all pieces be 3D printed, so I could build the tools I needed to complete the project myself. It took thousands of hours of work, and the ability to bring together multiple fields into complex solutions that spanned engineering, science and computer science.”
Apart from the technical problems associated with his project, Minnick faced challenges stemming from his lack of soldering experience and his access only to tools he could find in his garage.
“My initial ineptitude cost time and money, destroyed electronics, and yes, even set a printer on fire,” said Minnick. “Finding ways to work around my own circumstances and limitations posed serious problems. However, this would reveal itself to be a blessing in disguise. Interestingly, my lack of formal engineering training allowed me to ask questions and propose solutions that were novel.
“For instance, it was thought that 3D printers required microprocessors to function, so attempts to have a printer self-replicate centered around allowing it to etch circuit boards, but it could never actually manufacture the microelectronics themselves. As I gained more experience with 3D printers, my lack of formal training and experience led me to the creative and sometimes absurd solutions which sidestepped these difficult problems.”
Additive manufacturing captured Minnick’s imagination when he saw his first 3D printer during a visit to the University of Maryland.
“I immediately knew that I was hooked for life,” expressed Minnick.
The inspiration kept on coming. “My machine learning class inspired parts of the data strip generator. 19th-century street organs inspired the data strip system. A print-in-place toy robot—one of my first ever prints—inspired the print-in-place kinematics used for self-assembly.
“It was synthesizing solutions from all of these unique areas, making mistakes, and finding new and creative solutions to these problems, that drove my fascination with this device. While school teaches us to be perfect, real innovation is anything but. This project gave me a new outlook on problems that will carry over to all aspects of my life.”
Minnick believes wholeheartedly that in-person STEM events are essential for interacting with finalists from around the globe.
“The personal connections are not something that can be replicated virtually at all,” conveyed Minnick. “I have the benefit and the curse of having attended an in-person ISEF event in 2019. It was one of the best experiences I had, and it is the reason why my project stands as it does today—the event inspired me to continue developing my project.”
Minnick contends that web platforms like ProjectBoard can augment the in-person experience to make research more available to a wider audience outside of the competition. (Full disclosure: ProjectBoard is developed and owned by engineering.com.)
“One benefit of the virtual competition was that I could read about the other finalists’ work on ProjectBoard,” said Minnick. “In-person, this is something we did not really have time for—but my favorite part of the research process is sharing and learning from other researchers, so this was very nice. I actually think that incorporating the ProjectBoard system into a regular, in-person ISEF would be a fantastic addition. It would allow competitors and even those interested in exploring the projects (from outside the immediate area of the convention) to see the research.”
When asked if he had words of wisdom for future generations of STEM students, Minnick said: “Without failure, you cannot succeed. Every firmware bug was a search you will not have to make in the future, every machine crash a mistake you will not make again. Fail quickly and learn.”
To read more about Minnick’s project, click here.
An AI Model for Drought Stress Assessment in Plants
With his project, John Estrada was seeking to predict drought stress in bell peppers using AI and his custom-built robotic camera.
“Drought impacts 40 percent of the world’s population and is the most serious threat to crops in nearly every part of the world, especially in California,” stated Estrada. “It induces stress in plants, which negatively impacts crop yield. I witnessed the damage brought about by prolonged drought in my agricultural community in California’s Central Valley, and it inspired me to help our farmers find a solution to this devastating problem.”
According to Estrada, the commonly used Crop Water Stress Index (CWSI) is an old and complex calculation that uses variables which are mostly indirect indicators of drought stress. To help farmers conserve water and optimize yield, Estrada created an Artificial Intelligence Drought Assessment (AIDA) model for predicting drought stress quickly and accurately in bell pepper plants using variables that are all direct stress indicators. He developed the AIDA model using RGB light reflectance values, radiometric infrared values and soil moisture data.
Estrada programmed his model on a Raspberry Pi 3B using Python 3.7, utilizing Google’s TensorFlow 2.0 AI platform, and the Keras API. A sequential model with three neural net layers was constructed using Adam as the optimizer and mean squared error as the loss function. Metrics used in the model were mean squared error and mean absolute error in order to formulate an AI model that tracked closely with CWSI. An early stopping function was incorporated to prevent overfitting. Estrada trained his model from scratch, since there were no predetermined and validated models available for the determination of plant drought stress.
“By determining the appropriate weight of each variable in the prediction of the CWSI value, I was able to derive a very robust and accurate AIDA model,” said Estrada. “Although my model was trained using indoor data from a controlled environment, it can still be considered valuable because it can be applied in the field. To predict drought stress under field conditions, my AIDA model must be re-trained using data obtained from field experiments. This is not difficult to do since it can adapt easily to different conditions. The re-training phase will involve only one extra step that will take a few minutes.”
COVID-19 lockdowns prevented Estrada from performing his experiments under field conditions.
“I was supposed to use my custom-built Unmanned Aerial Vehicle (UAV) or drone equipped with an RGB and IR camera to gather field data in developing my AIDA model,” said Estrada. “Due to the pandemic, I had to conduct my experiment indoors.”
Estrada went on to build a robotic camera that could take both RGB and thermal images to obtain the pixel-by-pixel light reflectance values and canopy temperature readings needed to develop his AIDA model. The robotic arm had six degrees of freedom for maneuvering attachments around bell pepper plants. Estrada then improvised a way to attach both an RGB action camera and the infrared camera that he had built to the tip of the arm, using a camera case along with cable ties.
“It had to be sturdy enough to hold the two in place without moving, and yet not restrict the movements of the robotic arm,” described Estrada. “Then I programmed the arm movements around the plants so that photographs and radiometric images could be taken at specific X-Y-Z coordinates consistently.”
One trend that Estrada observed while reviewing his pair plots was the positive correlation between CWSI and green light reflectance values. When looking for explanations, he discovered the stay-green phenomenon in bell pepper plants—where plants affected by lack of water were becoming greener, rather than adhering to the opposite expectation.
“I almost cried for joy,” expressed Estrada. “This phenomenon was able to provide my model the possibility of predicting the earliest signs of drought stress that the CWSI could not do. It was a serendipitous moment showing how science is usually very deliberate, but sometimes unexpected findings can occur that you must be ready to explore and use to your advantage.”
Estrada attributes his project success to his programming and coding skills, as well as his ability to think outside the box. Next, he plans to conduct field experiments and adapt his AIDA model accordingly.
“Apart from farmers using my sensors to predict the earliest signs of drought stress and take appropriate intervention measures, plant researchers can also use my model to conduct phenotyping on drought resistant plants,” said Estrada. “Commercial and residential landscapes can potentially use the AIDA model to make sure that water is used judiciously. My dream is to integrate my AIDA model in the irrigation system and develop an app that is available for farmers around the globe. Hopefully, my technology will help in ending food insecurity—especially in developing countries.”
While Estrada missed the human connection associated with in-person events, he was positive about his virtual ISEF experience.
“It was personally a great experience for me,” said Estrada. “I had a lot of fun attending many events hosted during the week. I think the virtual platform was very fun to play around in and explore. ProjectBoard was easy to set up, and I enjoyed its look and presentation. It felt almost like a virtual reality world, which I thought was really cool.”
Estrada believes that one of the major benefits of holding ISEF virtually is the ease of joining the competition without the hassle of travelling, allowing more students to participate. As such, he foresees a hybrid model for future STEM events.
Estrada’s final words of encouragement to STEM students: “Keep on persevering, and do not be scared of sailing into unexplored waters. Think original and out of the box. You never know, you might discover a solution to a problem in your community that has never been thought of. Also, work on your communication skills. The only time you can say your project is a success, is when you are able to communicate your results effectively to the scientific community and the public.”
To read more about Estrada’s project, click here.
Read more from today’s generation and ISEF’S winners, here.
Engineering.com would like to congratulate both Minnick and Estrada for their inspiring work in the world of science. To explore Regeneron ISEF 2021’s full list of award-winning projects, click here.