Scilab and Finite element solvers conjunction
Paul Carrico, Esterline
Auxitrol SA (based in Bourges – France) is the French site of Esterline Advanced Sensor platform which is specialized in design and manufacturing sensors for aerospace capabilities include pressure, temperature, mass flow, torque, speed measurements.
At the design stage of any of our sensors, numerical tools are intensively used, and Scilab is part of them. We are using Scilab to perform different types of “classical” scientific calculations and test analyses:
- Strain-stress calculations from strain gauge measurements after vibration tests: calculation of the equivalent stress versus the frequency for example – preparation of numerical Finite Elements Analyses (FEA’s) and optimization process,
- Thermal/fluid calculations such as drag pressure/force/coefficient, Reynolds number, Nusselt one, heat exchange coefficients, vortex shedding frequencies, etc. the previous data are used to model our sensors such as Thermocouples, Cryogenic sensors, fuel/oil thermal ones, or bleed-air sensors:
- Automatic analyses of mechanical test results for materials characterization (from tensile tests to determine parameters such as Young modulus, Yield stress, Ultimate tensile stress),
Scilab has become an essential tool in our processes, especially in conjunction with our Finite Elements solvers (Samcef®, Radioss®) through:
- Building of parametric models of sensors in an early stage: geometry, material features, boundary conditions for example, in preamble to optimization steps,
- Realization of complex geometries: the high capabilities of both the pre-processing module of the solver and the Scilab ones are exploited in order to introduce variables in the models,
- Optimization in a broad sense, for parameters fitting, to find the best dimensions, to reduce the stress level, to readjust the model to the dynamic/vibration behavior of our sensor, to avoid forbidden frequencies provided by the customer, etc.
- In the previous item, Scilab drives the solver i.e. writes the input files, launches the different modules of the solver, records the values, calculates the cost function, exchanges with the optimizer (Scilab itself or Dakota libraries from the Sandia National Laboratories), updates the variables and restarts the new loop up to the final optimized values,
- FEA’s “runs” management in order to schedule the calculations, and to optimize the simulation resources (devices, licenses)
- Development of a fatigue post-processing module is in progress (critical plane approach, Neuber plasticity correction…)
The methodology and the tools have been successfully used on the development of Airbus A350 XWB sensors: Scilab became the skeleton of our numerical modeling activities…
Présentation non-disponible pour des raisons de confidentialité