More Like this

1. Inverse Design Framework With Invertible Neural Networks for Passive Vibration Suppression in Phononic Structures

   https://doi.org/10.1115/1.4052300  

   Oddiraju, Manaswin ; Behjat, Amir ; Nouh, Mostafa ; Chowdhury, Souma ( February 2022 , Journal of Mechanical Design)

   null (Ed.)

   Abstract Automated inverse design methods are critical to the development of metamaterial systems that exhibit special user-demanded properties. While machine learning approaches represent an emerging paradigm in the design of metamaterial structures, the ability to retrieve inverse designs on-demand remains lacking. Such an ability can be useful in accelerating optimization-based inverse design processes. This paper develops an inverse design framework that provides this capability through the novel usage of invertible neural networks (INNs). We exploit an INN architecture that can be trained to perform forward prediction over a set of high-fidelity samples and automatically learns the reverse mapping with guaranteed invertibility. We apply this INN for modeling the frequency response of periodic and aperiodic phononic structures, with the performance demonstrated on vibration suppression of drill pipes. Training and testing samples are generated by employing a transfer matrix method. The INN models provide competitive forward and inverse prediction performance compared to typical deep neural networks (DNNs). These INN models are used to retrieve approximate inverse designs for a queried non-resonant frequency range; the inverse designs are then used to initialize a constrained gradient-based optimization process to find a more accurate inverse design that also minimizes mass. The INN-initialized optimizations are found to be generally superior in terms of the queried property and mass compared to randomly initialized and inverse DNN-initialized optimizations. Particle swarm optimization with INN-derived initial points is then found to provide even better solutions, especially for the higher-dimensional aperiodic structures. 

   more » « less
2. Bacterial community dynamics during embryonic development of the little skate (Leucoraja erinacea)

   https://doi.org/10.1186/s42523-021-00136-x  

   Mika, Katelyn ; Okamoto, Alexander S. ; Shubin, Neil H. ; Mark Welch, David B. ( October 2021 , Animal Microbiome)

   Microbial transmission from parent to offspring is hypothesized to be widespread in vertebrates. However, evidence for this is limited as many evolutionarily important clades remain unexamined. There is currently no data on the microbiota associated with any Chondrichthyan species during embryonic development, despite the global distribution, ecological importance, and phylogenetic position of this clade. In this study, we take the first steps towards filling this gap by investigating the microbiota associated with embryonic development in the little skate,Leucoraja erinacea,a common North Atlantic species and popular system for chondrichthyan biology.

   To assess the potential for bacterial transmission in an oviparous chondrichthyan, we used 16S rRNA amplicon sequencing to characterize the microbial communities associated with the skin, gill, and egg capsule of the little skate,at six points during ontogeny. Community composition was analyzed using the QIIME2 pipeline and microbial continuity between stages was tracked using FEAST.

   We identify site-specific and stage-specific microbiota dominated by the bacterial phylaProteobacteriaandBacteroidetes. This composition is similar to, but distinct from, that of previously published data on the adult microbiota of other chondrichthyan species. Our data reveal that the skate egg capsule harbors a highly diverse bacterial community–particularly on the internal surface of the capsule–and facilitates intergenerational microbial transfer to the offspring. Embryonic skin and external gill tissues host similar bacterial communities; the skin and gill communities later diverge as the internal gills and skin denticles develop.

   Our study is the first exploration of the chondrichthyan microbiota throughout ontogeny and provides the first evidence of vertical transmission in this group.

    

   more » « less
3. VAX: Using Existing Video and Audio-based Activity Recognition Models to Bootstrap Privacy-Sensitive Sensors

   https://doi.org/10.1145/3610907  

   Patidar, Prasoon ; Goel, Mayank ; Agarwal, Yuvraj ( September 2023 , Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies)

   The use of audio and video modalities for Human Activity Recognition (HAR) is common, given the richness of the data and the availability of pre-trained ML models using a large corpus of labeled training data. However, audio and video sensors also lead to significant consumer privacy concerns. Researchers have thus explored alternate modalities that are less privacy-invasive such as mmWave doppler radars, IMUs, motion sensors. However, the key limitation of these approaches is that most of them do not readily generalize across environments and require significant in-situ training data. Recent work has proposed cross-modality transfer learning approaches to alleviate the lack of trained labeled data with some success. In this paper, we generalize this concept to create a novel system called VAX (Video/Audio to 'X'), where training labels acquired from existing Video/Audio ML models are used to train ML models for a wide range of 'X' privacy-sensitive sensors. Notably, in VAX, once the ML models for the privacy-sensitive sensors are trained, with little to no user involvement, the Audio/Video sensors can be removed altogether to protect the user's privacy better. We built and deployed VAX in ten participants' homes while they performed 17 common activities of daily living. Our evaluation results show that after training, VAX can use its onboard camera and microphone to detect approximately 15 out of 17 activities with an average accuracy of 90%. For these activities that can be detected using a camera and a microphone, VAX trains a per-home model for the privacy-preserving sensors. These models (average accuracy = 84%) require no in-situ user input. In addition, when VAX is augmented with just one labeled instance for the activities not detected by the VAX A/V pipeline (~2 out of 17), it can detect all 17 activities with an average accuracy of 84%. Our results show that VAX is significantly better than a baseline supervised-learning approach of using one labeled instance per activity in each home (average accuracy of 79%) since VAX reduces the user burden of providing activity labels by 8x (~2 labels vs. 17 labels).

    

   more » « less
4. The Green’s Function-Based Thermal Analysis of a Spherical Geothermal Tank in a Semi-Infinite Domain

   https://doi.org/10.1115/1.4054568  

   Wang, Tengxiang ; Wu, Chunlin ; Zhang, Liangliang ; Yin, Huiming ( July 2022 , Journal of Applied Mechanics)

   Abstract The Green’s function of a bimaterial infinite domain with a plane interface is applied to thermal analysis of a spherical underground heat storage tank. The heat transfer from a spherical source is derived from the integral of the Green’s function over the spherical domain. Because the thermal conductivity of the tank is generally different from soil, the Eshelby’s equivalent inclusion method (EIM) is used to simulate the thermal conductivity mismatch of the tank from the soil. For simplicity, the ground with an approximately uniform temperature on the surface is simulated by a bimaterial infinite domain, which is perfectly conductive above the ground. The heat conduction in the ground is investigated for two scenarios: First, a steady-state uniform heat flux from surface into the ground is considered, and the heat flux is disturbed by the existence of the tank due to the conductivity mismatch. A prescribed temperature gradient, or an eigen-temperature gradient, is introduced to investigate the local temperature field in the neighborhood of the tank. Second, when a temperature difference exists between the water in the tank and soil, the heat transfer between the tank and soil depends on the tank size, conductivity, and temperature difference, which provide a guideline for heat exchange design for the tank size. The modeling framework can be extended to two-dimensional cases, periodic, or transient heat transfer problems for geothermal well operations. The corresponding Green’s functions are provided for those applications. 

   more » « less
5. Adaptive flow-induced vibration control using distributed sensor/actuator networks with gated recurrent units and genetic algorithms

   https://doi.org/10.1117/12.2630805  

   Abdelaziz, Khalid M. ; Hobeck, Jared D. ( April 2022 , SPIE 12046, Sensors and Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems 2022)

   Zonta, Daniele ; Su, Zhongqing ; Glisic, Branko (Ed.)

   This work presents a control scheme for wind-induced vibration mitigation for tall buildings based on a gated recurrent unit (GRU) encoder-decoder model which operates using readings from multiple sensors to define a unique system state. The sensors include a distributed network of pressure probes installed on surrounding buildings, accelerometers installed on the principal building, and atmospheric conditions. The encoder-decoder GRU is trained from timeseries sensor readings to construct a unique internal representation (hidden state) of the evolving wind and building conditions. A 1:400-scale aeroelastic building model with motorized plates acting as aerodynamic control surfaces is used in wind tunnel experiments to conduct this study. An online genetic reinforcement learning (GRL) algorithm uses a series of multilayer perceptron (MLP) networks to determine optimum actuator orientations for different flow conditions. The algorithm stores previously discovered solutions in the MLPs sorted by their fitness. The GA operates by obtaining a solution from each of the MLPs and performing GA operations on them to choose the next combination of plate angles to try. A chance also exists for trying completely random plate angles to prevent the GA from stalling. The MLPs are continuously trained during online optimization using findings obtained from new trials. The system eliminates the need for holding wind conditions, which are uncontrollable, constant during online training but still uses a pseudo-random search technique to obtain global optimum solutions. Results show a considerable reduction in building RMS acceleration when compared with a large collection of results with random constant plate angle orientations. 

   more » « less