Generative prediction of laser-induced rocket ignition with dynamic latent space representations

Tony Zahtila; Ettore Saetta; Murray Cutforth; Davy Brouzet; Diego Rossinelli; Gianluca Iaccarino

cs.LGMar 8, 2026

Generative prediction of laser-induced rocket ignition with dynamic latent space representations

Authors: Tony Zahtila, Ettore Saetta, Murray Cutforth, Davy Brouzet, Diego Rossinelli, Gianluca Iaccarino

Organizations: Center for Turbulence Research, Stanford University, USA · University of Naples Federico II, Italy

Abstract

Accurate and predictive scale-resolving simulations of laser-ignited rocket engines are highly time-consuming because the problem includes turbulent fuel-oxidizer mixing dynamics, laser-induced energy deposition, and high-speed flame growth. This is conflated with the large design space primarily corresponding to the laser operating conditions and target location. To enable rapid exploration and uncertainty quantification, we propose a data-driven surrogate modeling approach that combines convolutional autoencoders (cAEs) with neural ordinary differential equations (neural ODEs). The present target application of an ML-based surrogate model to leading-edge multi-physics turbulence simulation is part of a paradigm shift in the deployment of surrogate models towards increasing real-world complexity. Sequentially, the cAE spatially compresses high-dimensional flow fields into a low-dimensional latent space, wherein the system's temporal dynamics are learned via neural ODEs. Once trained, the model generates fast spatiotemporal predictions from initial conditions and specified operating inputs. By learning a surrogate to replace the entirety of the time-evolving simulation, the cost of predicting an ignition trial is reduced by several orders of magnitude, allowing efficient exploration of the input parameter space. Further, as the current framework yields a spatiotemporal field prediction, appraisal of the model output's physical grounding is more tractable. This approach marks a significant step toward real-time digital twins for laser-ignited rocket combustors and represents surrogate modeling in a complex system context.

Figures & tables

Figure 1: Workflow describing the construction and deployment of the neural ODE and decoder as a generative model.

Figure 2: Ignition probability density functions (PDFs) marginalized over the input parameters $\xi_{0}$ through $\xi_{14}$ , comparing ignition success (red) and ignition failure (blue) through kernel density estimates; green check marks denote the five parameters with the largest KL divergence between igniting and non-igniting cases. Full description of parameters in Table 2 in Appendix A .

Figure 3: Quantity of interest: spatiotemporal fields, (a) rendered simulation output of the successfully ignited combustor, (b) illustration of the integrated quantity acquisition, (1) light beam generation, (2) ray evolution according to the Eikonal equation , (3) Snell’s law and occlusion, and (4) irradiance sampling, (c) successful and unsuccessful time-series for the computational infrared, and (d) computational Schlieren.

Figure 4: Architecture schematic. (a) Encoder: a sequence of convolutional, with residual connections and pooling blocks maps the single–channel input to progressively smaller but deeper feature maps, which are then flattened and passed through fully-connected layers to obtain an 8-dimensional latent state

\mathbf{V}

. (b) NODE: the latent state is augmented with the control inputs (

\xi,t

), with two fully connected layers and the latent vector

\mathbf{V}

is forecast for the duration of the ignition trial. (c) Decoder: the latent state

\mathbf{V}

is lifted back to high dimension by fully-connected layers and a sequence of convolutional and upsampling blocks reconstructing the output field.

Figure 5: (a) Evolution of the validation loss against epoch during the training of the AE, for choices of latent-space dimension hyperparameter $N_{\ell}=1,4,8$ and $16$ . (b) Representative loss during training of the NODE for the 200,000 iterations, for choice of hidden layer width hyperparameter $w=400$ . The horizontal black line indicates the loss criterion for a curriculum learning increment, and markers at spike locations denote curriculum updates.

Figure 6: Representative case of latent-space trajectories ${V}_{i}$ illustrating the curriculum learning process. Panels (a)–(d) display the advancing training window (gray band), highlighting the progressive learning of newly revealed trajectory segments as the time horizon extends. Dashed lines indicate the beginning prediction at the start of the update, while solid lines show the final learned trajectory.

Figure 7: Latent variables, up-sampled in time, for the reacting rocket combustor LES ensemble, post-processed with coloring by a component of the input vector, $\xi_{2}=\beta=R_{1}/R_{2}$ .

Figure 8: Latent-space projections of $V_{5}$ against all other latent variables for the reacting sparks simulation ensemble, colored by the input parameter $\xi_{2}=\beta=R_{1}/R_{2}$ . Each trajectory begins at a green hollow circle and terminates at a red hollow circle, illustrating consistent trajectories in the low-dimensional manifold and separation of igniting from non-igniting cases.

Figure 9: Training (left) and validation (right) trajectories generated by the neural ODE, overlaid on the latent space representation learned by the AE for predictions with RMSE error of

5^{th}

50^{th}

, and

95^{th}

percentiles indicating strong, median and weak prediction. Ensemble distribution for entire latent space trajectories also shaded.

Metric	Train				Validation
	Q1	Q2	Q3	Q4	Q1	Q2	Q3	Q4
Fréchet Distance	0.009	0.013	0.018	0.035	0.075	0.100	0.190	0.289
L2 Norm	0.002	0.003	0.005	0.007	0.021	0.029	0.048	0.080

Table 1: Error metric in NODE-predicted latent space trajectories across four temporal quarters (Q1–Q4) across train-validation data split. Results reflect transition from early-time kernel dynamics to late-time divergence in the chaotic reacting phase

Figure 10: Confusion matrices for the binary classification of ignition outcomes. Left panel shows the training set results, with near-perfect separation between failure and success cases. Right panel shows the validation set performance.

Figure 11: One-to-one comparison of DnAE reconstructions with CFD reference sequences for representative ignition outcomes corresponding to the confusion-matrix categories in Figure 10 . (a) Training ignition success, (b) training non-igniting case, (c) correctly predicted non-ignition, (d) spurious ignition prediction, (e) missed ignition prediction, and (f) correctly predicted ignition.

Figure 12: Examples of generative latent-space trajectory predictions for unseen cases, classified as either igniting or non-igniting. For reference, range of trajectories on the low-dimensional manifold extracted from compressed CFD data are shaded in background, where $\hat{V}_{f}$ denotes latent trajectories with ignition failure and $\hat{V}_{s}$ for success.

Figure 13: Joint ignition probability distributions, estimated via Monte Carlo sampling of the NODE latent space trajectories, as a function of the parameter pairs (

\beta,E

) (top row) and (

\beta,l_{axial}

) (bottom row), obtained using, (a) 300 samples, (b) 200,000 samples, (c) 500,000 samples, and (d) 1,000,000 samples. Additionally marked are decision boundaries for

Pr(ign)=0.5,0.75,0.9

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Uncertainty ID	Description	Experiments	Distribution
$\xi_{0}\!:\Delta x_{l}$	Streamwise focal imprecision	Y	$N\sim(0.29\,\text{mm},0.04\,\text{mm}^{2})$
$\xi_{1}\!:\Delta x_{r}$	Radial focal location	Y	$N\sim(-0.54\,\text{mm},0.20\,\text{mm}^{2})$
$\xi_{2}\!:\beta=R1/R2$	Lobe radii ratio		$U\sim[1.1,2.1]$
$\xi_{3}\!:\alpha=l_{axial}/2R1$	Aspect ratio		$U\sim[2.0,2.5]$
$\xi_{4}\!:l_{axial}$	Laser axial length		$U\sim[1.44\text{mm},2.16\text{mm}]$
$\xi_{5}\!:E$	Energy deposited	Y	$U\sim[20\,\text{mJ},54\,\text{mJ}]$

Appendix

Table 2: Summary of the uncertainty parameters,

\xi_{i}

, of the laser pulse for the ignition trials, including their descriptions and associated probability distributions. Probability distributions are denoted as

U\sim[a,b]

(uniform probability between

a

and

b

) or

N\sim(\mu,\sigma^{2})

(normal distribution with mean

\mu

and variance

\sigma^{2}

Neural Surrogate Modeling

Neural surrogate modeling uses learned models to approximate costly simulations, enabling faster prediction, design, control, and inference.

See all 410 papers

Mar 16, 2026

A convolutional autoencoder and neural ODE surrogate modeling framework applied to transient counterflow flames

A novel convolutional autoencoder and neural ODE (CAE-NODE) framework is proposed for a reduced-order model (ROM) applied to transient 2D counterflow flames, as an extension of AE-NODE methods in homogeneous reactive systems to spatially resolved flows. The multidimensional thermochemical fields (256 x 256 grid, 21 variables) obtained from direct numerical simulations (DNS) are used in training the CAE, where convolutional layers learned the underlying spatial correlations, allowing the CAE to construct an unsupervised 3D latent manifold that is physically meaningful, smooth, and continuous in time. This results in a compression ratio of over 400,000 times. The NODE then subsequently learns the continuous-time dynamics on the latent manifold, enabling the prediction of the full temporal evolution of the flames by integrating forward in time from an initial condition. The results demonstrate that the CAE-NODE can accurately capture the entire transient process, including ignition, flame propagation, and the gradual transition to a non-premixed condition, with excellent agreement with the DNS, while adhering to conservation principles at virtually no computational cost compared to the reference DNS. Predictions remain accurate at strain rates outside the training range. Moreover, despite being unsupervised, the learned latent manifold is highly correlated with the flame-state descriptors such as the progress variable, mixture fraction, and the scalar dissipation rate. This study, for the first time, highlights the potential of CAE-NODE for surrogate modeling of unsteady dynamics of multi-dimensional reacting flows.

Mert Yakup Baykan, Weitao Liu, Mohammad Rafi Malik +4

Department of Aerospace Engineering, KAIST, Daejeon, Republic of Korea · Clean Energy Research Platform (CERP), KAUST, Thuwal, Saudi Arabia · Institute for Reactive Flows (IRST), University of Stuttgart, Pfaffenwaldring 31, 70569 Stuttgart, Germany

Autoencoders PDE Surrogate Modeling Convolutional Autoencoder Reduced-Order Modeling Neural ODEs Computational Fluid Dynamics

Ask PDF

Jul 5, 2026

A Deep Learning-based surrogate model for Severe Accidents in nuclear reactors using ASTEC

Integral codes like the Accident Source Term Evaluation Code (ASTEC) are powerful tools to study the physics of Severe Accidents (SAs) in nuclear reactors. Real time SA simulators can also be helpful in training operators of nuclear plants to react correctly to malfunctions. However, SA simulators can take up to several days per simulation, making their use infeasible for real time applications. In this work we show how to speed up a SA simulator with a fast, Deep Learning based (DL), surrogate model (SM). The SM is built as a combination of a dimensionality reduction stage, via an AutoEncoder, and a time-stepping stage, via a Neural Ordinary Differential Equation. The data on which the SM is trained are obtained from the ASTEC simulator, by sampling a set of operator actions for station blackout (SBO) and loss-of-coolant accidents (LOCA). The objective of the developed SM is to approximate multiple spatio-temporal fields for the thermal-hydraulic physics, core degradation, and fission product release modules in ASTEC's vessel domain. The SM predicts simultaneously around

80

different physical variables (both scalar and fields), maintaining a stable autoregressive rollout up to

50

thousand time steps. In addition, the AutoEncoder achieves a dimensionality reduction by a factor of over

300

, which allows the SM to predict up to

40

hours of simulation in under a minute, both on CPU and GPU. This work is the first study of the capabilities and limits of DL based surrogate modeling in approximating the challenging, highly non-linear physics of ASTEC.

Alessandro Longhi, Danny Lathouwers, Zoltán Perkó

TU Delft · Department of Radiation Science and Technology · Delft, The Netherlands

Neural Surrogate Modeling Surrogate Modeling Physics-Based Simulation

Ask PDF

Aug 18, 2026

Adaptive surrogate modeling for high-dimensional spatio-temporal output

This paper develops an adaptive surrogate modeling method for problems with very high-dimensional spatio-temporal outputs. The analysis of spatio-temporal multi-physics systems is computationally expensive and consists of a large number of inputs and outputs. Surrogate models are often constructed to replace the physics-based model to achieve computational efficiency in analyses such as uncertainty quantification and optimization that require many function calls. In order to address the challenge introduced by the high dimensionality of spatio-temporal output, a dimension reduction method is first employed to map the high-dimensional output to a low-dimensional latent space. This is followed by the construction of the surrogate model in the low-dimensional space. The prediction error in the original space, which includes both the reconstruction error and surrogate model error, is evaluated using different error metrics. Based on the prediction accuracy of the surrogate model, new training points are identified for adaptive improvement of the surrogate model. We present a novel adaptive sampling technique that combines exploration and exploitation to improve the surrogate model accuracy with the fewest possible runs of the expensive physics-based model. Thermo-mechanical analysis of a gas turbine engine blade is used to analyze the effectiveness of the proposed method.

Berkcan Kapusuzoglu, Shunsaku Matsumoto, Yoshitomo Miyagi +2

Department of Civil and Environmental Engineering, Vanderbilt University, Nashville, TN 37235, USA · Strength Research Department, Research and Innovation Center, Mitsubishi Heavy Industries, Ltd., Nagasaki, 851-0392, Japan · Strength Research Department, Research and Innovation Center, Mitsubishi Heavy Industries, Ltd., Takasago, 676-8686, Japan

Surrogate Modeling Adaptive Sampling

Ask PDF

Abstract

Neural Surrogate Modeling

Explore similar work

A convolutional autoencoder and neural ODE surrogate modeling framework applied to transient counterflow flames

A Deep Learning-based surrogate model for Severe Accidents in nuclear reactors using ASTEC

Adaptive surrogate modeling for high-dimensional spatio-temporal output