"ZOOM" SEMINAR

SCHOOL OF MECHANICAL ENGINEERING SEMINAR
Monday, December 30, 2020 at 14:00

Modelling, Prediction and Mitigation of Nonlinear Complex Dynamical Systems Under Stochastic Disturbances

Dr. Maor Farid

Massachusetts Institute of Technology, Department of Mechanical Engineering, 77 Massachusetts Ave., Cambridge, MA 02139, USA

Technion-Israel Institute of Technology, Department of Mechanical Engineering, Technicon City, Haifa 3200, Israel

faridm@mit.edu

Abstract. Destructive vibrations caused by stochastic external disturbances such as earthquakes, explosions, and fluid-structure interactions pose an insidious threat to various critical infrastructure and mechanical components. Among the most vulnerable and potentially hazardous are large structures, nuclear facilities, and tanks containing toxic fluids—all of which are prevalent in the chemical, nuclear, and petroleum industries. Damage to these structures not only endangers human lives, but also often result in long-term environmental and ecological distress. Hence, reliable models that predict the dynamical responses and resulting stresses from stochastic excitations in real-time are a critical necessity for the operation, maintenance and structural health monitoring of these systems. However, most of the aforementioned systems consist of numerous smaller elements of various shapes and physical properties—hence they are referred to as complex systems—which are a great challenge to model and predict. Therefore, the vast majority of existing prediction methods are based on either over-simplistic models that suffer from lack of reliability, or time-consuming computational models that are not applicable for real-time purposes and do not describe the underlying mechanisms that govern the system’s dynamical response.

The current research focuses on a multidisciplinary approach, combining both reduced-order modelling and machine-learning-based methods for description and prediction of complex dynamical systems. Finally, hybrid vibration mitigation technologies are introduced, which utilize both linear and nonlinear energy absorption techniques for enhanced protection performances, and broader energy and frequency ranges of effectiveness. The methods presented here pave the way towards faster and more accurate forecasting methods as well as more effective vibration mitigation technologies, allowing for enhanced resistance of existing facilities, structures, and mechanical components to a variety of external disturbances in real-time. Thus, they are of great importance in terms of public health, homeland security, and environmental responsibility.

Bio: Dr. Maor Farid is a Fulbright and ISEF postdoctoral fellow at the Stochastic Analysis and Nonlinear Dynamics Lab at MIT and the Nonlinear Dynamical Systems Lab at the Technion, and a fellow of the postdoctoral programs of the Israel Academy of Sciences and Humanities and the Shamir program of the Israeli Ministry of Science, Technology and Space. Farid serves as an adjunct lecturer in the Department of Mechanical Engineering at the Technion, and currently offers a course he developed, “Analytic and Machine Learning-Based Modeling of Dynamical Systems.” He obtained his B.Sc. (2013), M.Sc. (2015), and Ph.D. (2017) in Mechanical Engineering from the Technion. He completed his Ph.D. at 24 as the youngest graduate at the Technion, in parallel to his mandatory military service in the Brakim excellence program, where he served as an R&D engineer in the Israeli Prime Minister's Office, and as a commander and an AI researcher in the Data Science team of Unit 8200. He was recognized as an “Excellent Scientist” (top 3 in the military) and completed his service with the rank of captain. Farid served as an adjunct lecturer at Ben-Gurion University in a course he developed based on his Ph.D. research and practical experience in the Israeli defense industry, called “Modeling and Analysis of Dynamical Systems.” He is the director of the Israeli scientists’ center of ScienceAbroad at MIT, and the founder and CEO of Lilmod Lehatsliach (Learn to Succeed), a non-profit organization that empowers at-risk youth and undergraduates from the Israeli periphery. Among other awards, he is the recipient of the Moskowitz and "Spirit of Zion" Awards for Zionism, leadership, and social activism, and was included in the Forbes 30 Under 30 list. In 2020, Dr. Farid was elected for the leadership program of Harvard Kennedy School for emerging and promising leaders. His research interests center on nonlinear dynamics, stochasticity, chaos theory, and machine learning.

https://zoom.us/j/96584758181?pwd=WC9PMXdsYzJ3NFdEN2Q5ZUtOZEVjdz09 The meeting will be recorded and made available on the School’s site.

~~

"ZOOM" SEMINAR
SCHOOL OF MECHANICAL ENGINEERING SEMINAR
Wednesday, January 6, 2021 at 14:00
InverSynth 2: Deep End-to-End Synthesizer Parameters Inference
Or Rimoch,
Mechanical Engineering department
Under the supervision of Dr. Noam Koengstein, Industrial Engineering department

Predicting Synthesizer parameters configuration, given it's raw audio or spectrogram is a challenging task to achieve. This subject has enormous implications for both the music industry and professional recordings studios. Current predictive methods are based on auto-regressive and generative models, such Autoencoders and GANS as the main ways to tackle this problem. In this work, we use Encoder-Decoder network architecture, termed InverSynth 2. Differ from previous work, we combined it with a unique multi-loss that has two loss terms: Cross Entropy (CE) and L2. The CE aim is to solve a classification task, by predicting the synthesizer parameters, as the output of the Encoder, and the L2 loss aim to solve a regression task, by predicting 2D spectrogram (STFT), as the output of the Decoder. Given a raw audio signal the network is able to predict the synthesizer configuration in an End-to-End fashion: The ground truth Spectrogram is fed into the Encoder to get the predicted parameters of the Synthesizer, then it fed as input to the Decoder, to get the predicted spectrogram. this way, the Decoder serves as a proxy for a "real" synthesizer, except that it outputs a spectrogram instead of raw audio. by using this technique, we can get richer representation feedback loop, while training the model
 In our results, we analyze, qualitatively and quantitatively, two synthesizer datasets: "synthetic" dataset - based on Python code that emulates a Synthesizer, and "real" dataset - based on real VST plugin called "TAL-NoiseMaker".
This work is still under ongoing research, and further improvements regarding network architecture and metrics might take place.

Join Zoom Meeting

https://zoom.us/j/96584758181?pwd=WC9PMXdsYzJ3NFdEN2Q5ZUtOZEVjdz09 The meeting will be recorded and made available on the School’s site.