Brite/EuRam project BE963046
Partners: CRF (FIAT), SNECMA, Politecnico di Torino, CNRS, IntraSoft
Industry is urgently engaged in improving its competitive position: shortening time to market, cost reduction, product quality improvement. Computer-aided engineering (CAE) occupies a strategic position through its support for design optimisation and reduction of costly physical experimentation. Using CAE for multi-criteria, global design optimisation paves the way for the construction of a truly Concurrent Engineering environment.
High computational effort is a barrier to successful and fast optimisation, vital for Concurrent Engineering. Typical practice involves running simulation software in a sequential, manual, fashion with local, piecemeal optimisation, leading at best to local optima.
These limitations are made more serious by the increased complexity of design problems, the interactions between different domains (mechanical, thermal, electronic etc.) and the large range of available simulators.
The urgent need is to achieve the following objectives:
- to integrate two or more simulators within the design optimisation process;
- to develop strategies for drastic reduction of the computational effort needed to optimise a design, based on current state-of-the-art multi--objective optimisation and global algorithms.
The CE2 project
The CE2 project starts by constructing mathematical functions, emulators, that represent simulator code. Thus an emulator is a highly adaptive but still simple mathematical model, which mimics the simulator over the whole design space. It is built by experimenting on the simulator using a well-chosen set of inputs. Different emulators can be combined for global optimisation, multi-objective optimisation, sensitivity analysis and fast graphical display. This substitution of the simulator represents a trade-off between accuracy and fast analysis.
The project will culminate in a decision support system that incorporates state-of-the-art multi-objective algorithms on a supporting foundation of statistical emulators, derived from the design and analysis of computer experiments conducted on the simulators.
It is expected that the software will be ready for industrial use at the end of the project in 1999 with full commercial partnerships available.
State-of-the-Art and Degree of Innovation
The use of statistical emulators in computer aided design dates from 1989, and is now a proven and fast-growing area on the frontier between statistical modelling and CAE. The main advantages are that the emulators can run hundreds or thousands of times faster than the simulators and hence can be used for sophisticated analysis which the relative slowness of a standard simulator can prevent.
Main advantages are:
- Sensitivity analysis and robust engineering design (RED)
- Exploration of the CAE output response surface as a function of inputs, based on a choice of state-of-the-art emulators
- Inverse engineering and target attainment
- Global optimisation
- Complex and multi-objective criteria Extension to any energy domain
The CE2 project incorporates extensive case--studies with major partners from the automotive and aerospace industries across several domains simultaneously, including
- Aircraft jet engine design
- Automobile suspension design
- Automobile powertrain design
The CE2 project will be benchmarked by comparison with other systems.
Key to the successful implementation of CE2 is the automation of the emulator building phase. During the development of CE2 the team will be seeking enhancement in a number of ways which are not part of the main project brief. It invites partnerships with CAD/CAE software companies (and users) to develop connectors to the CE2 project software.
In particular the integration with feature--based design, fast auto--meshing, computational geometry, are all open areas for joint research and commercial development.
Home/ The project/ The partners/ Feed-back