About

Steve Brunton is a Professor in Mechanical Engineering at the University of Washington, where he holds the Boeing Professor in AI & Data-Driven Engineering position. He is also a Data Science Fellow at the eScience Institute. His research areas include data science & artificial intelligence, fluids and thermal sciences, and advanced manufacturing. Brunton's work focuses on applying data-driven methods and artificial intelligence to engineering problems, particularly in the context of energy, manufacturing, and fluid dynamics. His contributions involve integrating data science techniques with traditional engineering disciplines to advance understanding and innovation in these fields.

Research topics

• Computer Science
• Artificial Intelligence
• Machine Learning
• Mathematics
• Mechanics
• Engineering
• Data Mining
• Theoretical computer science
• Algorithm
• Aerospace engineering
• Geometry
• Physics
• Engineering ethics
• Software engineering
• Data science
• Programming language

Selected publications

• Decoding complexity through machine learning is redefining scientific discovery

  Communications Physics · 2026-05-14

  articleOpen accessSenior author

  Abstract As scientific instruments and the literature generate ever larger volumes of data, machine learning (ML) has become essential for organizing, analyzing and interpreting complex information. This Perspective examines how ML accelerates discovery across disciplines, with examples such as brain mapping and exoplanet detection. It also considers situations with different levels of prior knowledge about the underlying phenomenon, outlining strategies to address limitations and exploit ML effectively. Although growing reliance on ML raises challenges for research practice and validation, it is reshaping scientific methods and expanding what can be studied. We also highlight foundation models as a promising route to faster, broader scientific discovery.

  PublisherOA PDFDOI

• Computed Tomography Using Meta-Optics

  ACS Photonics · 2025-03-01 · 3 citations

  article

  Computer vision tasks require processing large amounts of data to perform image classification, segmentation, and feature extraction. Optical preprocessors can potentially reduce the number of floating-point operations required by computer vision tasks, enabling low-power and low-latency operation. However, existing optical preprocessors are mostly learned and hence strongly depend on the training data and thus lack universal applicability. In this paper, we present a meta-optic imager, which implements the Radon transform, obviating the need for training the optics. High-quality image reconstruction with a large compression ratio of 9.2% is presented through the use of the simultaneous algebraic reconstruction technique. We also demonstrate image classification with 90% accuracy on a further compressed (0.6% of total measured pixels) Radon data set through a neural network trained on digitally transformed images. Our work shows the efficacy of data-independent encoding in an optical encoder. While our platform is based on meta-optics, we note that such encoding can be performed with other optics as well.

  PublisherDOI

• A data-driven approach for predicting the cetane number of renewable and unconventional fuels

  Energy Conversion and Management · 2025-12-29

  articleOpen access

  Recent research on engine combustion has focused on new less carbon-intensive alternatives that meet current standards and can serve as drop-in fuels. However, developing surrogate fuels is challenging due to the difficulty of predicting key properties such as derived cetane number, which quantifies the ignition quality in compression ignition engines. Higher values indicate highly reactive fuels whereas lower values indicates longer ignition times that will lead to misfiring and increased emissions in diesel engines. The measurement of this property entails the test of the candidate fuels under specific conditions, making it costly and time-consuming. This work explores the use of data-driven models to predict the derived cetane number of both single-component fuels and blends. NREL’s compendium is used as reference, while the ignition quality tester is replaced by a 1D spray model coupled with detailed chemical kinetics. The models are trained on fuel composition, basic thermophysical properties, and simulated ignition delay. Additionally, symbolic regression is employed to derive an explicit equation linking derived cetane number to key fuel descriptors. The resulting expression achieves comparable accuracy to state-of-the-art neural networks (MAE 3.5, R 2 0.96–0.97), but reveals a physically interpretable structure where ignition delay dominates predictions, with double bond equivalent and heat of vaporization acting as corrective terms for specific chemical families, enhancing generalizability across diverse fuels. The accuracy of the resulting models is then compared to the available literature. This research provides a robust framework for predicting derived cetane number in unconventional fuels, enabling the exploration of new candidates for on-road, off-road, and aviation applications. • Developed multiple data-driven models to predict the DCN of unconventional fuels. • NREL’s DCN compendium is used as a reference combined with a 1D spray model. • An expression derived through symbolic regression links DCN to key fuel descriptors. • Explicit equation predicts DCN with high accuracy (MAE 3.5) and interpretability.

  PublisherDOI

• Controlling chaotic energy events in fluids with reinforcement learning

  2025-06-16

  articleOpen accessSenior author

  Extreme energy events arise spontaneously in turbulent flow, and often lead to catastrophic outcomes of the system and its surrounding environment.This work focuses on controlling extreme events in the canonical Kolmogorov flow, which is a sinusoidally-driven, two-dimensional turbulent flow that exhibits extreme energy dissipation events due to non-linear energy transfers at different scales.

  PublisherOA PDFDOI

• Physics-inspired data-driven modeling of complex mechanical components in mecha(tro)nic systems

  Mechanical Systems and Signal Processing · 2025-08-29

  articleSenior author
  PublisherDOI

• Reduced-order modeling and classification of hydrodynamic pattern formation in gravure printing

  Machine Learning Engineering · 2025-08-21

  articleOpen access

  Hydrodynamic pattern formation phenomena in printing and coating processes are still not fully understood.However, fundamental understanding is essential to achieve high-quality printed products and to tune printed patterns according to the needs of a specific application like printed electronics, graphical printing, or biomedical printing.The aim of the paper is to develop an automated pattern classification algorithm based on methods from supervised machine learning and reduced-order modeling.We use the HYPA-p dataset, a large image dataset of gravure-printed images, which shows various types of hydrodynamic pattern formation phenomena.It enables the correlation of printing process parameters and resulting printed patterns for the first time.A total of 26 880 images of the HYPA-p dataset have been labeled by a human observer as dot patterns, mixed patterns, or finger patterns; 864 000 images (97%) are unlabeled.A singular value decomposition is used to find the modes of the labeled images and to reduce the dimensionality of the full dataset by truncation and projection.Selected machine learning classification techniques are trained on the reduced-order data.We investigate the effect of several factors, including classifier choice, whether or not fast Fourier transform (FFT) is used to preprocess the labeled images, data balancing, and data normalization.The best performing model is a k-nearest neighbor (kNN) classifier trained on unbalanced, FFT-transformed data with a test error of 3%, which outperforms a human observer by 7%.Data balancing slightly increases the test error of the kNN-model to 5%, but also increases the recall of the mixed class from 90% to 94%.Finally, we demonstrate how the trained models can be used to predict the pattern class of unlabeled images and how the predictions can be correlated to the printing process parameters, in the form of regime maps.

  PublisherOA PDFDOI

• Separation of periodic orbits in the delay-embedded space of chaotic attractors

  Proceedings of the Royal Society A Mathematical Physical and Engineering Sciences · 2025-09-01 · 1 citations

  articleOpen accessSenior author

  This work explores the intersection of time-delay embeddings, periodic orbit theory and symbolic dynamics. Time-delay embeddings have been effectively applied to chaotic time-series data, offering a principled method to reconstruct relevant information of the full attractor from partial time-series observations. In this study, we investigate the structure of the unstable periodic orbits (UPOs) of an attractor using time-delay embeddings. First, we embed time-series data from a periodic orbit into a higher-dimensional space through the construction of a Hankel matrix, composed of time-shifted copies of the data. We then examine the influence of the width and height of the Hankel matrix on the geometry of UPOs in the delay-embedded space. The right singular vectors of the Hankel matrix provide a basis for embedding the periodic orbits. We observe that increasing the height of the Hankel matrix leads to a clear separation of the periodic orbits into distinct clusters within the embedded space. Our analysis characterizes these separated clusters and provides a mathematical framework to determine the relative position of individual UPOs in the embedded space. In addition, we present a modified formula to derive symbolic representation of distinct periodic orbits for a specified sequence length, extending the Redfield–Pólya (R–P) enumeration theorem.

  PublisherDOI

• A deep learning approach to wall-shear stress quantification: from numerical training to zero-shot experimental application

  Journal of Fluid Mechanics · 2025-06-25 · 4 citations

  preprintOpen access

  The accurate quantification of wall-shear stress dynamics is of substantial importance for various applications in fundamental and applied research, spanning areas from human health to aircraft design and optimization. Despite significant progress in experimental measurement techniques and postprocessing algorithms, temporally resolved wall-shear stress fields with adequate spatial resolution and within a suitable spatial domain remain an elusive goal. Furthermore, there is a systematic lack of universal models that can accurately replicate the instantaneous wall-shear stress dynamics in numerical simulations of multiscale systems where direct numerical simulations (DNSs) are prohibitively expensive. To address these gaps, we introduce a deep learning architecture that ingests wall-parallel streamwise velocity fields at $y^+ \approx 3.9 \sqrt {Re_\tau }$ of turbulent wall-bounded flows and outputs the corresponding two-dimensional streamwise wall-shear stress fields with identical spatial resolution and domain size. From a physical perspective, our framework acts as a surrogate model encapsulating the various mechanisms through which highly energetic outer-layer flow structures influence the governing wall-shear stress dynamics. The network is trained in a supervised fashion on a unified dataset comprising DNSs of statistically one-dimensional turbulent channel and spatially developing turbulent boundary layer flows at friction Reynolds numbers ranging from $390$ to $1500$ . We demonstrate a zero-shot applicability to experimental velocity fields obtained from particle image velocimetry measurements and verify the physical accuracy of the wall-shear stress estimates with synchronized wall-shear stress measurements using the micro-pillar shear-stress sensor for Reynolds numbers up to $2000$ . In summary, the presented framework lays the groundwork for extracting inaccessible experimental wall-shear stress information from readily available velocity measurements and thus, facilitates advancements in a variety of experimental applications.

  PublisherOA PDFDOI

• Author response: Sensorimotor delays constrain robust locomotion in a 3D kinematic model of fly walking

  2025-03-20

  peer-reviewOpen access

  Walking animals must maintain stability in the presence of external perturbations, despite significant temporal delays in neural signaling and muscle actuation. Here, we develop a 3D kinematic model with a layered control architecture to investigate how sensorimotor delays constrain robustness of walking behavior in the fruit fly, Drosophila. Motivated by the anatomical architecture of insect locomotor control circuits, our model consists of three component layers: a neural network that generates realistic 3D joint kinematics for each leg, an optimal controller that executes the joint kinematics while accounting for delays, and an inter-leg coordinator. The model generates realistic simulated walking that resembles real fly walking kinematics and sustains walking even when subjected to unexpected perturbations, generalizing beyond its training data. However, we found that the model’s robustness to perturbations deteriorates when sensorimotor delay parameters exceed the physiological range. These results suggest that fly sensorimotor control circuits operate close to the temporal limit at which they can detect and respond to external perturbations. More broadly, we show how a modular, layered model architecture can be used to investigate physiological constraints on animal behavior.

  PublisherOA PDFDOI

• Accelerating scientific discovery with the common task framework

  ArXiv.org · 2025-11-06

  preprintOpen accessSenior author

  Machine learning (ML) and artificial intelligence (AI) algorithms are transforming and empowering the characterization and control of dynamic systems in the engineering, physical, and biological sciences. These emerging modeling paradigms require comparative metrics to evaluate a diverse set of scientific objectives, including forecasting, state reconstruction, generalization, and control, while also considering limited data scenarios and noisy measurements. We introduce a common task framework (CTF) for science and engineering, which features a growing collection of challenge data sets with a diverse set of practical and common objectives. The CTF is a critically enabling technology that has contributed to the rapid advance of ML/AI algorithms in traditional applications such as speech recognition, language processing, and computer vision. There is a critical need for the objective metrics of a CTF to compare the diverse algorithms being rapidly developed and deployed in practice today across science and engineering.

  PublisherOA PDFDOI

Recent grants

• AI Institute in Dynamic Systems

  NSF · $19.8M · 2021–2026

Frequent coauthors

• J. Nathan Kutz

  289 shared

• Joshua L. Proctor

  Seattle University

  81 shared

• Bernd R. Noack

  79 shared

• Eurika Kaiser

  60 shared

• Bingni W. Brunton

  University of Washington

  57 shared

• J. Nathan Kutz

  39 shared

• Krithika Manohar

  33 shared

• Benjamin Strom

  University of Washington

  33 shared

Education

• PhD, Mechanical and Aerospace Engineering

  Princeton University

  2012

• BS, Mathematics

  California Institute of Technology

  2006

Awards & honors

• Data Driven Science and Engineering: Machine Learning, Dynam…
• Brunton, Noack, Koumoutsakos. Machine Learning for Fluid Mec…
• Brunton, Proctor, Kutz. Discovering governing equations from…
• Kutz, Brunton, Brunton, Proctor. Dynamic Mode Decomposition:…
• Brunton & Noack. Closed-loop turbulence control: Progress an…