The central obstacle the researchers set out to attack is not a lack of algorithms but a lack of data. Every candidate design in a supersonic flow problem must be evaluated with computational fluid dynamics simulations that resolve unsteady, compressible, reacting flow, and each of those simulations can take substantial computing time and resources. Reinforcement learning, which typically improves by trial and error across thousands or millions of attempts, is notoriously hungry for exactly this kind of feedback. When each trial costs a full simulation of a shock-laden supersonic shear layer, the standard playbook collapses under its own appetite. The team, led by Meng You, Tingting Liu, Tianzi Bai, Shuqin Jia, and Ying Huai, therefore framed their contribution around sample efficiency: extracting the maximum design insight from the smallest possible number of flow simulations.

Their solution combines two ideas. The first is a deep policy gradient network, a class of reinforcement learning agent that learns a probability distribution over actions and nudges that distribution toward actions that yield better outcomes. In this application, the actions are the parameters of a pulsed fuel injection scheme: the pulsed frequency, the pulsed amplitude, and the mean total pressure of the injected jet. The second idea is an adaptive noise-scaling framework governing how the agent explores. In policy gradient methods, randomness in the action selection is what allows the agent to discover unfamiliar regions of the design space, but too much noise wastes evaluations on poor candidates, while too little traps the agent in mediocre local optima. The adaptive scheme dynamically adjusts the scale of this exploration noise, balancing exploration against exploitation as the optimization proceeds.

A distinctive feature of the architecture is the role of a predictor deep neural network. Rather than requiring a full flow simulation for every candidate the policy network considers, the predictor provides fast approximate feedback on how a given combination of pulse frequency, amplitude, and mean total pressure is likely to perform. This internal surrogate lets the policy network adjust its parameters toward the desired mixing performance without paying the full simulation cost at every step. The predictor effectively acts as a cheap stand-in for the physics, absorbing much of the trial-and-error burden, while the true simulations are reserved for verifying and calibrating the most promising candidates. The result is an optimization loop in which expensive computational fluid dynamics evaluations are spent sparingly and deliberately.

The physical system the team used as its proving ground is a chemical laser, specifically a supersonic flow device in which mixing quality is measured by the small-signal gain coefficient, a quantity that directly determines laser performance. In a chemical oxygen-iodine laser, reactive streams must mix rapidly at supersonic speed for the lasing reaction to proceed efficiently, making the small-signal gain coefficient an unusually sensitive and practically meaningful yardstick for mixing. Pulsed injection itself is a well-established mixing enhancement technique: by modulating the injected jet in time, engineers can excite flow structures that stir the streams far more effectively than a steady jet of the same average strength. The trick has always been choosing the right modulation parameters, a search space large enough that manual tuning or exhaustive sweeps quickly become impractical.

The numerical foundation of the study was established with careful verification. The team constructed three computational grids of increasing refinement, containing 60,000, 120,000, and 540,000 cells respectively, and compared Mach number distributions along the centerline for a representative operating condition with a pulse frequency of 100 kilohertz and a mean total pressure of 250 Torr. The distributions from the two finer grids overlapped almost completely, indicating that 120,000 cells were sufficient to capture the physics. A parallel time-step independence study tested four temporal resolutions, from one-fiftieth to one-four-hundredth of the pulse period, and found that the time-averaged small-signal gain coefficient changed by less than two percent between the two finest steps. On this basis, the researchers adopted the medium grid with the T/200 time step, a pragmatic compromise that kept each simulation affordable without sacrificing accuracy.

With the simulation framework validated, the optimization proceeded on two fronts. The team demonstrated both single-objective and multi-objective optimization of the pulsed injection parameters, the latter addressing the reality that mixing performance often involves trade-offs among competing metrics that cannot all be maximized simultaneously. The adaptive noise-scaling sampling strategy proved decisive: compared with a deep policy gradient network using a conventional sampling strategy, the new framework achieved both superior accuracy and superior stability in locating optimal injection settings. Stability matters as much as accuracy in this context, because an optimizer that lands on a different answer every run is of little use to an engineer who needs a trustworthy design point.

The ultimate test came from verifying the optimized parameters against full simulations of supersonic reactive flows in the chemical laser. The relative errors of the small-signal gain coefficient predicted by the optimization framework, compared with the simulation results, remained below five percent across the cases examined. That level of agreement indicates that the learned policy and its predictor network were not merely exploiting artifacts of the training data but had genuinely captured the relationship between injection parameters and mixing performance. For a quantity as sensitive as small-signal gain in a reacting supersonic flow, sub-five-percent error from a data-efficient learning framework is a substantial result, and it suggests the approach could transfer to other high-speed flow applications where the same trade-off between data cost and optimization quality exists.

The implications reach well beyond chemical lasers. Scramjet engines, which hold fuel and air at supersonic speeds throughout combustion, face perhaps the most demanding mixing problem in propulsion, and pulsed fuel injection has attracted growing attention there, including recent studies of hydrogen-fueled and kerosene-fueled supersonic combustors. Supersonic ejectors and other high-speed flow devices share the same fundamental challenge. In all of these fields, the bottleneck has been the same: the simulations needed to evaluate a design are expensive, so the space of possible designs goes largely unexplored. A framework that demonstrably reduces the data burden while preserving optimization accuracy changes the economics of that exploration, potentially allowing engineers to survey design spaces that were previously out of reach.

The study also sits within a broader movement in fluid dynamics toward machine-learning-driven flow control. Recent years have seen deep reinforcement learning applied to active control of turbulent separation bubbles, jet mixing optimization using Bayesian methods and evolutionary algorithms, and machine learning approaches to scramjet combustor configuration. What distinguishes the Dalian work is its explicit focus on the sample-efficiency problem that has limited practical adoption, and its demonstration on a real, industrially relevant quantity, the gain of a chemical laser, rather than a purely academic benchmark. By coupling a policy gradient agent with adaptive noise scaling and a learned predictor, the researchers offer a template that other groups working on data-limited optimization problems, from propulsion to materials design, may find adaptable. The work was supported by the Chinese Academy of Sciences and the Dalian National Laboratory for Clean Energy, and the authors state they have no competing interests. As hypersonic flight and directed-energy systems mature, tools that squeeze better performance out of every costly simulation are likely to become as valuable as the physics they help to optimize.

Subject of Research: Deep reinforcement learning optimization of pulsed injection parameters for enhancing supersonic mixing in high-speed flow devices such as chemical lasers and scramjets

Article Title: Optimization of Pulsed Injection for Enhancing Supersonic Mixing via Deep Reinforcement Learning with Adaptive Noise Scaling

Article References: You, M., Liu, T., Bai, T., Jia, S., & Huai, Y. (2026). Optimization of Pulsed Injection for Enhancing Supersonic Mixing via Deep Reinforcement Learning with Adaptive Noise Scaling. International Journal of Aeronautical and Space Sciences. https://doi.org/10.1007/s42405-026-01246-z

Image Credits: AI Generated

DOI: 10.1007/s42405-026-01246-z

Keywords: supersonic mixing, deep reinforcement learning, pulsed injection, policy gradient network, adaptive noise scaling, chemical laser, scramjet, computational fluid dynamics, small-signal gain coefficient, sample efficiency, flow control, multi-objective optimization

The research, led by Xiaolong Zhang of Nanjing University of Aeronautics and Astronautics together with colleagues including Hao Wang of Rongtong Aeroengine Technology and corresponding author Ning Ge, tackles a problem that has challenged engineers since the earliest scramjet experiments of the 1990s. When a shock wave strikes a boundary layer—the thin layer of slowed air hugging a surface—it can thicken that layer dramatically, or even force it to separate from the wall entirely. For film cooling, this is a double-edged sword: a thicker boundary layer might seem to offer more insulation, but separation bubbles create recirculating hot gas that can strip away the coolant film and expose the wall to searing mainstream temperatures. Understanding precisely where and how this happens is essential for designing cooling schemes that survive in hypersonic propulsion systems.

To capture these interactions, the team employed a hybrid turbulence modeling strategy called the Stress-Blended Eddy Simulation, or SBES, coupled with a fifth-order weighted essentially non-oscillatory scheme known as WENO-ZQ. The approach is a compromise between computational necessity and physical fidelity. Full large-eddy simulation, which resolves nearly all turbulent eddies, is extraordinarily expensive for the complex geometries and high Reynolds numbers of engine-relevant flows. Standard Reynolds-averaged methods, meanwhile, average away the very unsteady structures—vortex rings, shear-layer instabilities, separation bubbles—that govern film cooling performance. SBES blends the two: it runs in RANS mode near the wall where turbulence is dominated by small-scale stresses, and switches smoothly to LES mode in the outer regions where large, energy-containing eddies matter. The fifth-order WENO-ZQ scheme, developed originally by Zhu and Qiu, provides the shock-capturing sharpness needed to resolve discontinuities without spurious oscillations.

The validation of this numerical toolkit was itself a significant contribution. The researchers demonstrated that the SST–SBES combination could accurately reproduce the vortex structures and their temporal evolution within the film cooling domain—structures such as the counter-rotating vortex pair that lifts coolant off the wall, the horseshoe vortices that wrap around the jet as it emerges, and the hairpin vortices that trail downstream. These vortical structures are the machinery of mixing: they determine whether the cool air stays pressed against the wall where it protects the surface, or gets churned into the hot mainstream where it is wasted. A turbulence model that cannot capture them cannot be trusted to predict cooling effectiveness, which is precisely why the team invested effort in establishing the method’s resolution and accuracy before drawing physical conclusions.

The geometry at the heart of the study is the 7-7-7 shaped hole, a design whose cross-section expands laterally and vertically in a stepped fashion, resembling a dustpan laid into the surface. Shaped holes of this kind were developed to combat the fundamental weakness of cylindrical holes: the strong counter-rotating vortex pair that lifts the coolant jet away from the wall. By diffusing the jet exit, the shaped hole reduces the jet’s vertical momentum and spreads the coolant laterally, improving coverage. But shaped holes were largely characterized in subsonic conditions. What happens when a shock wave impinges upstream of such a hole, in a supersonic mainstream, remained poorly quantified—especially at different blowing ratios, the ratio of coolant momentum to mainstream momentum.

The team simulated a supersonic flat-plate configuration with the dustpan-shaped hole, examining blowing ratios of 0.5 and 0.65. The first major finding concerns the shock–boundary layer interaction itself. When the shock strikes the incoming boundary layer upstream of the hole, it thickens the layer and can induce a separation bubble—a pocket of recirculating flow detached from the wall. Interestingly, the blowing ratio, which changes the character of the jet emerging from the hole, had only a limited effect on the mainstream pressure distribution. In other words, the shock system upstream behaves largely as if the cooling jet were not there, decoupling the external aerodynamics from the internal cooling problem in a way that simplifies how engineers can think about the two phenomena.

The second key finding is more subtle and concerns the region near the hole’s trailing edge. At the lower blowing ratio of 0.5, the jet does not fill the expanded dustpan exit uniformly. Instead, aerodynamic bulges form within the jet—localized protrusions of high-speed coolant that act like small compression surfaces, generating compression waves of their own. These waves, the study found, intensify localized boundary layer thickening near the trailing edge of the hole, compounding the disturbance already created by the upstream shock. At the higher blowing ratio of 0.65, however, the jet has enough momentum to fill the expanded exit more completely, smoothing out the bulges and alleviating this additional thickening. The practical implication is striking: simply increasing the coolant supply can suppress a secondary shock-generating mechanism, though at the cost of more compressed air diverted from the compressor.

Perhaps the most elegant results come from spectral analysis—the turbulence equivalent of listening to the flow’s frequency signature. The researchers placed monitoring points upstream, midstream, and downstream of the interaction and examined the power spectra of the pressure fluctuations. Upstream and at midstream locations, the spectra displayed a distinctive −7/3 power-law scaling, a slope steeper than the classical Kolmogorov −5/3 law of inertial-range turbulence. The team attributes this anomalous scaling to two mechanisms working together: the compression imposed by the shock wave, which compresses turbulent eddies and steepens their spectra, and the shear-layer instability driven by the interaction between the jet and the mainstream. The −7/3 slope has precedent in pressure spectra of turbulent free shear flows, and its appearance here ties the shock-perturbed cooling flow into a broader family of compressible turbulent phenomena.

Downstream, however, the story changes. The spectra at far-field monitoring points transition toward the classical −5/3 inertial subrange slope, indicating that the violent, shock-dominated disturbances have decayed and the flow has relaxed into a state where turbulent mixing between the jet and the mainstream has saturated. In physical terms, the shock’s fingerprint fades with distance, and the jet-mainstream system settles into ordinary turbulent mixing. This spectral evolution gives engineers a diagnostic tool: by measuring pressure fluctuations at a surface, one can infer how far downstream the shock’s influence on the cooling film persists, and where the protective layer has either recovered or been fully consumed by mixing.

The study, funded by China’s National Science and Technology Major Project, arrives at a moment when hypersonic flight programs worldwide are pushing film cooling technology into regimes where shock interactions are unavoidable rather than exceptional. In a scramjet combustor, shock trains and expansion fans are everywhere, and cooling holes sit directly in their path. The finding that higher blowing ratios can suppress jet-induced compression waves offers a concrete design lever, while the demonstrated fidelity of the SBES–WENO-ZQ framework provides a validated virtual testbed for exploring hole geometries, spacing, and coolant conditions without costly wind-tunnel campaigns. As shaped holes migrate from subsonic turbines into supersonic combustors, studies like this one are building the quantitative foundation that turns a fragile cool blanket into engineering that can be trusted at Mach 5 and beyond.

Subject of Research: Shock wave interaction with shaped-hole film cooling in supersonic boundary layers

Article Title: SBES-Based Study on Aerodynamic Impacts of Upstream Shock Waves on Film Cooling from Dustpan-Shaped Holes

Article References: Zhang, X., Wang, H., Wu, X., & Ge, N. (2026). SBES-Based Study on Aerodynamic Impacts of Upstream Shock Waves on Film Cooling from Dustpan-Shaped Holes. International Journal of Aeronautical and Space Sciences. https://doi.org/10.1007/s42405-026-01272-x

Image Credits: AI Generated

DOI: 10.1007/s42405-026-01272-x

Keywords: film cooling, shock wave, SBES, dustpan-shaped hole, supersonic flow, boundary layer separation, turbulence modeling, WENO scheme, spectral analysis, scramjet, gas turbine, aerodynamics