Multi-objective optimization for ship hull form design.
NUMECA is now Cadence
The unique optimization software for your CAE environment
Optimizing your design starts here.
Fidelity™ Optimization is a unique optimization framework allowing for multidisciplinary, multiobjective design optimization - all based on the highly efficient DoE generation and surrogate modeling of Minamo.
With the added option of uncertainty quantification, engineers can design under real-world conditions, taking into account variations on input conditions, geometry or a combination, and ensuring performance within the design envelope.
Connected to all Cadence software solvers, and open to other software codes, Fidelity™ Optimization can be customized for easy integration into the design process.
See the full workflow in action in this video.
Smart parameterization is key to the success of optimization.
AxCent™ and Autoblade™ provide parameterization environments dedicated to turbomachinery. They offer flexible modeling to define a large design space and can easily handle all types of turbomachinery blading systems, with classical design models based on blade sections, camber, thickness, and stacking.
For marine design projects, Cadence has developed a direct plug-in with the Grasshopper geometry modeler, enabling naval architects and marine engineers to easily create 3D geometries.
Morphing and adjoint solvers
A powerful alternative to a dedicated parameterization tool, morphing method allows engineers to modify, or deform, a non-parameterized geometry or mesh.
Arbitrary changes in the body shape can be achieved without changing the mesh topology: control points become the parameters of the optimization.
All the features for efficient exploration of the design space can be achieved with Minamo's DoE and optimization algorithms, or by coupling to an adjoint solver.
Optimizing the Design of Experiment
The optimization kernel Minamo is developed by Cenaero and based on the Design of Experiment (DoE) technique.
The way the design space is explored has a huge impact on the convergence of the optimization and on the search for the optimum value.
Minamo includes superior space-filling techniques that greatly reduce the number of samples needed for a design of experiments.
Auto-adaptive DoE techniques are also available for efficient handling of high-dimensional design spaces, as well as highly constrained optimization problems.
A task manager enables spreading the CFD runs on a cluster, for the generation of multiple samples in parallel.
Optimization and surrogate modeling with Minamo
The surrogate model rapidly speeds up exploration of the design space by approximating the solution in between the DoE points of the sampled design space.
Several algorithms are available, from Artificial Neural Networks to Radial Basis Function Networks (RBF) and Kriging. The surrogate model is then used to apply an iterative optimization algorithm and find the global optimum.
This approach produces benefits at different stages of the development process, from the early concept definition to the final design phase, and speeds up the development cycle of complex products by reducing computational costs related to high-fidelity simulations.
Data-mining & analysis
Analysis of Variance (ANOVA)
Data-mining algorithms are used to get a deeper understanding of the design space and to provide more insights during the optimization process, thus reducing the CPU cost of the overall optimization.
Based on the surrogate model, Analysis of Variance (ANOVA) provides information on which design variables have more influence over the outcome.
Self Organizing Maps project high-dimensional data onto a low-dimensional (typically 2D) map. These visual inspection tools are useful for revealing major trends and discovering correlations and anti-correlations among parameters and/or responses.
Multidisciplinary design optimization
An optimization process is stronger if it can account for off-design conditions, mechanical reliability, noise, and other aspects such as manufacturing costs. Fidelity™ Optimization has been developed in order to allow for that flexibility, with the possibility to call a flexible amount of evaluators and solvers.
All the solvers, CFD, acoustics, and structural can be coupled and included within the analysis phase, with a flexible open framework, where users can plug in all Cadence and external solvers to benefit from state-of-the-art technology.
Additionally users can optimize not only the design performance, but ensure and control operating range by including multiple operating conditions in the analysis.
Uncertainty Quantification and Robust Design Optimization
CFD and design procedures have always been applied assuming geometries and operational conditions are well known. In reality they are subject to manufacturing variability and variable operating conditions.
Uncertainty quantification allows to acount for these uncertainties and estimates their impact on system performance. It can be used to identify variables with the largest sensitivity for instance, or calculate the impact of less stringent manufacturing tolerances.
In design mode, uncertainty quantification enables minimizing the impact of uncertainties on performance, ensuring that the final design performance is robust. Through this Robust Design Optimization (RDO) the performance will be less sensitive to the variability of those uncertain variables.
Resulting marine propeller shapes comparing standard vs robust design optimization.
Fidelity™ Optimization is used to optimize the aerodynamics performances of a centrifugal compressor wheel.
Multidisciplinary multipoint robust design optimization of turbocharger impeller, a case from FORD Motor.
- Multidisciplinary optimization platform
- All Cadence tools for mesh generation, CFD analysis and post-processing are plugged-in
- FEA/CFD analysis
- Turbomachinery and Marine dedicated workflow
- Generic workflows for multiphysics optimization
- Parametric modelers such as Autoblade™, Axcent™, Rhino, Omnis™/Morphing, user-defined functions or tools
- Minamo: modern and highly efficient optimization kernel
- Latest techniques for DoE generation: Latin hypercube sampling, Voronoi tessellation-based sampling etcetera
- Advanced surrogate models, such as Radial Basis Function Networks, Kriging, and others
- Single-objective or aggregate objective optimization algorithms
- Multi-objective algorithm (Pareto)
- Comprehensive data-mining and analysis tools, including Self-Organizing Maps, ANOVA, and more
- Legacy optimization kernel
- Numerous techniques for DoE generation: Latin hypercube, Rechtschaffner, Plackett-Burmann, D-optimal sampling and more
- Neural network assisted single and multi-objective optimization
- Unique Uncertainty Quantification for operational, geometrical and manufacturing variability
- Unique Robust Design Optimization