skip to main content

Title: 3D printable biomimetic rod with superior buckling resistance designed by machine learning

Our mother nature has been providing human beings with numerous resources to inspire from, in building a finer life. Particularly in structural design, plenteous notions are being drawn from nature in enhancing the structural capacity as well as the appearance of the structures. Here plant stems, roots and various other structures available in nature that exhibit better buckling resistance are mimicked and modeled by finite element analysis to create a training database. The finite element analysis is validated by uniaxial compression to buckling of 3D printed biomimetic rods using a polymeric ink. After feature identification, forward design and data filtering are conducted by machine learning to optimize the biomimetic rods. The results show that the machine learning designed rods have 150% better buckling resistance than all the rods in the training database, i.e., better than the nature’s counterparts. It is expected that this study opens up a new opportunity to design engineering rods or columns with superior buckling resistance such as in bridges, buildings, and truss structures.

Award ID(s):
Publication Date:
Journal Name:
Scientific Reports
Nature Publishing Group
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract Herein new lattice unit cells with buckling load 261–308% higher than the classical octet unit cell were reported. Lattice structures have been widely used in sandwich structures as lightweight core. While stretching dominated and bending dominated cells such as octahedron, tetrahedron and octet have been designed for lightweight structures, it is plausible that other cells exist which might perform better than the existing counterparts. Machine learning technique was used to discover new optimal unit cells. An 8-node cube containing a maximum of 27 elements, which extended into an eightfold unit cell, was taken as representative volume element (RVE). Numerous possible unit cells within the RVE were generated using permutations and combinations through MATLAB coding. Uniaxial compression tests using ANSYS were performed to form a dataset, which was used to train machine learning algorithms and form predictive model. The model was then used to further optimize the unit cells. A total of 20 optimal symmetric unit cells were predicted which showed 51–57% higher capacity than octet cell. Particularly, if the solid rods were replaced by porous biomimetic rods, an additional 130–160% increase in buckling resistance was achieved. Sandwich structures made of these 3D printed optimal symmetric unit cells showed 13–35%more »higher flexural strength than octet cell cored counterpart. This study opens up new opportunities to design high-performance sandwich structures.« less
  2. Abstract

    The brick-and-mortar structure inspired by nature, such as in nacre, is considered one of the most optimal designs for structural composites. Given the large number of design possibilities, extensive computational work is required to guide their manufacturing. Here, we propose a computational framework that combines statistical analysis and machine learning with finite element analysis to establish structure–property design strategies for brick-and-mortar composites. Approximately 20,000 models with different geometrical designs were categorized into good and bad based on their failure modes, with statistical analysis of the results used to find the importance of each feature. Aspect ratio of the bricks and horizontal mortar thickness were identified as the main influencing features. A decision tree machine learning model was then established to draw the boundaries of good design space. This approach might be used for the design of brick-and-mortar composites with improved mechanical properties.

  3. Cellular materials have been widely used in load carrying lightweight structures. Although lightweight increases natural frequency, low stiffness of cellular structures reduces natural frequency. Designing structures with higher natural frequency can usually avoid resonance. In addition, because of the less amount of materials used in cellular structures, the energy absorption capability usually decreases such as under impact loading. Therefore, designing cellular structures with higher natural frequency and higher energy absorption capability is highly desired. In this study, machine learning and novel inverse design techniques enable to search a huge space of unexplored structural designs. In this study, machine learning regression and Generative Neural Networks (GANs) were used to form an inverse design framework. Optimal cellular unit cells that surpass the performance of biomimetic structures inspired from honeycomb, plant stems and trabecular bone in terms of natural frequency and impact resistance were discovered using machine learning. The discovered optimal cellular unit cells exhibited 30–100% higher natural frequency and 300% higher energy absorption than those of the biomimetic counterparts. The discovered optimal unit cells were validated through experimental and simulation comparisons. The machine learning framework in this study would help in designing load carrying engineering structures with increased natural frequency and enhancedmore »energy absorption capability.« less
  4. Modern digital manufacturing processes, such as additive manufacturing, are cyber-physical in nature and utilize complex, process-specific simulations for both design and manufacturing. Although computational simulations can be used to optimize these complex processes, they can take hours or days--an unreasonable cost for engineering teams leveraging iterative design processes. Hence, more rapid computational methods are necessary in areas where computation time presents a limiting factor. When existing data from historical examples is plentiful and reliable, supervised machine learning can be used to create surrogate models that can be evaluated orders of magnitude more rapidly than comparable finite element approaches. However, for applications that necessitate computationally- intensive simulations, even generating the training data necessary to train a supervised machine learning model can pose a significant barrier. Unsupervised methods, such as physics- informed neural networks, offer a shortcut in cases where training data is scarce or prohibitive. These novel neural networks are trained without the use of potentially expensive labels. Instead, physical principles are encoded directly into the loss function. This method substantially reduces the time required to develop a training dataset, while still achieving the evaluation speed that is typical of supervised machine learning surrogate models. We propose a new method formore »stochastically training and testing a convolutional physics-informed neural network using the transient 3D heat equation- to model temperature throughout a solid object over time. We demonstrate this approach by applying it to a transient thermal analysis model of the powder bed fusion manufacturing process.« less
  5. Lisacek, Frederique (Ed.)
    Historically, the database search algorithms have been the de facto standard for inferring peptides from mass spectrometry (MS) data. Database search algorithms deduce peptides by transforming theoretical peptides into theoretical spectra and matching them to the experimental spectra. Heuristic similarity-scoring functions are used to match an experimental spectrum to a theoretical spectrum. However, the heuristic nature of the scoring functions and the simple transformation of the peptides into theoretical spectra, along with noisy mass spectra for the less abundant peptides, can introduce a cascade of inaccuracies. In this paper, we design and implement a Deep Cross-Modal Similarity Network called SpeCollate , which overcomes these inaccuracies by learning the similarity function between experimental spectra and peptides directly from the labeled MS data. SpeCollate transforms spectra and peptides into a shared Euclidean subspace by learning fixed size embeddings for both. Our proposed deep-learning network trains on sextuplets of positive and negative examples coupled with our custom-designed SNAP-loss function. Online hardest negative mining is used to select the appropriate negative examples for optimal training performance. We use 4.8 million sextuplets obtained from the NIST and MassIVE peptide libraries to train the network and demonstrate that for closed search, SpeCollate is able to performmore »better than Crux and MSFragger in terms of the number of peptide-spectrum matches (PSMs) and unique peptides identified under 1% FDR for real-world data. SpeCollate also identifies a large number of peptides not reported by either Crux or MSFragger. To the best of our knowledge, our proposed SpeCollate is the first deep-learning network that can determine the cross-modal similarity between peptides and mass-spectra for MS-based proteomics. We believe SpeCollate is significant progress towards developing machine-learning solutions for MS-based omics data analysis. SpeCollate is available at .« less