Matthieu Sacher

Associate Professor · Arts et Métiers Institute of Technology, École Navale, IRENav, France

I develop numerical and experimental methods for the design of lifting marine systems. My work combines fast flow models, fluid-structure interaction, experiments, and gradient-based or surrogate-based optimization.

Portrait of Matthieu Sacher

Research

Fluid-structure interaction

Fast nonlinear models for sails (vortex lattice, extended lifting line, semi-rigid sails and forestay) and for composite hydrofoils.

Shape optimization

Efficient global optimization (EGO) with Gaussian processes, multi-fidelity, constraints and non-computable domains; adjoint gradients by algorithmic differentiation.

Hydrofoils and cavitation

Panel methods (PUFFIn), free-surface effects, partial cavitation, robust section design.

Wind-assisted propulsion

Simultaneous design and routing of wind-assisted ships, representative weather conditions, collision risk.

Propulsors and oscillating foils

Experimental and multi-fidelity optimization of cycloidal propeller pitch laws; oscillating foils near the free surface.

Hydroacoustics

Noise radiated by lifting bodies and marine propulsors: Ffowcs Williams–Hawkings acoustic analogy coupled with unsteady panel or CFD flow solvers.

Software

XFOILAD

Gradient-based optimization of foil, wing and blade sections. The complete XFOIL solver, eN transition included, is differentiated in reverse mode with Tapenade: one gradient costs about one analysis plus one adjoint sweep, whatever the number of shape variables. Partial cavitation, robust design under uncertain turbulence and incidence, trailing-edge flap.

Foil section and flap schedule of a 50-knot foiling catamaran: section, pressure distributions and drag per speed
Quick start: main foil section of a 50-knot foiling catamaran, starting from a NACA 4410. The shape, one flap deflection per speed and the hinge position (19 variables) are optimized together to minimize drag at 50 kn, 35 kn and take-off (18 kn), with no cavitation at any speed and with thickness, bending stiffness and flap thickness constraints. Drag −13 % at 50 kn, −21 % at 35 kn and −20 % at take-off; about 2 min on a laptop.
Robust design over an uncertain inflow turbulence level: optimized sections, drag versus turbulence level and drag distributions
Robust design over an uncertain inflow turbulence level (Tu from 0.03 to 1 %, CL = 0.5, Re = 106), starting from a NACA 2412. Top: sections. (a) Drag versus Tu: each deterministic optimum is the best near its design turbulence level and among the worst elsewhere, while the robust sections (mean and CVaR0.9 of the drag) keep a long laminar run at every Tu. (b) Distribution of the drag: the CVaR optimum is nearly insensitive to turbulence.
Adjoint gradients with respect to every node coordinate compared with central finite differences
Gradient check: derivatives of lift, drag and incidence with respect to the y coordinate of every panel node (NACA 2412, Re = 106), adjoint (lines) against central finite differences (symbols). They agree to 10−4 or better, sharp peaks at the leading edge and transition points included, for the cost of one analysis plus one adjoint sweep.

PythonFortranGPL-3.0

Source code on GitHub

PUFFIn

Potential Unsteady code for Foils with Free surface INcluded: a three-dimensional unsteady boundary element (panel) method for lifting bodies such as hydrofoils, wings, rudders, stabilisers and fins. Bodies in an infinite fluid, near a wall or a free surface, fixed or with imposed motions, in regular or irregular waves; free wake, linearised free surface, viscous boundary layer with transition, sheet cavitation, dynamic positioning and hydroacoustics (Ffowcs Williams–Hawkings). A few seconds to a few minutes per computation on a laptop.

PUFFIn: Kelvin wave pattern generated by a hydrofoil moving under the free surface
Kelvin wave pattern generated by a hydrofoil moving beneath the free surface.
PUFFIn: three-blade vertical-axis propeller with the free wakes of its blades
Three-blade vertical-axis propeller computed with a free wake: vortex sheets shed by each blade over the rotation.

WindowsGraphical interfaceCommand linePython API

Developed with contributions from École Navale, ENSTA and Ifremer. Distributed as a ready-to-run Windows package with its user guide (no installation needed).

Download (link to come)

Publications

Loading…

Contact

Arts et Métiers Institute of Technology, École Navale, IRENav, France

matthieu.sacher@ensam.eu