Time to market reduction via statistical information management (TITOSIM)

Voice of Customer (VOC) methodologies consist in methods used in the first phase of the product and process development. The customer needs and requirements are collected and determined by specifically designed questionnaires, filled in by a sample of prospective customers. The data collected in this way is analysed, thus showing the customer demands. These are then translated into concrete targets for the designer. This phase is referred to as 'target setting'. KPA in particular has substantial experience in these areas and could contribute with the use of surveys, focus groups, pilot tests etc. The main methodology implemented in TITOSIM is the House of Quality module, consisting of an analysis tool and summarising graphics. The House of Quality, as an innovative tool, may be used in other sub-sequential phases of the development of a product. Consequently, the module may be used at different stages of the process. The House of Quality consists of different matrices that establish the correlation between customer requirements and technical requirements, technical and parts requirements, parts and process requirements and so on.

ROBUST DESIGN methodologies consist in methods for evaluating the impact of identified sources of noise on designs. New methodologies for Robust Design have been defined and integrated in the TITOSIM prototype. They enable the design engineer to specify tolerances on input factors, to encapsulate variation in component values, and also to specify target values of design performance that include variability, such as mean performance, variance of performance or more specific target distributions for performance values. Briefly, the TITOSIM methodologies are: 1.Stochastic responses (flexible definition of robust characteristics) 2.Stochastic emulators (efficient estimation of product and process variation in the presence of noise) 3.New robust design optimisation methods (Safe design and Minimax design) 4.Library of cost functions 5.Integration with VOC module There is a strong link between the Robust Design, VOC and optimisation modules. The VOC module identifies the input factors, classified as pure factors (no noise), internal noise factors, external noise factors. The Robust Design modules allow stochastic emulators to be built, based on stochastic responses. These emulators provide fast, efficient estimates of product and process variation that can be used in the optimization module to search for robust design solutions. To the knowledge of the partners this is a unique feature in TITOSIM.

One of the main objective of TITOSIM project is to develop a software tool (hereinafter referred as TITOSIM Prototype). Following the approach used in its predecessor (CE)2, this tool will not only help engineers to optimise real product designs by making efficient use of computer codes (simulators), but will also combine the use of statistical methods for the design and analysis of computer and physical experiments with algorithms for global and multiobjective optimisation to search for optimal design configurations, taking into account VOC. Regarding "computer experiments", TITOSIM will replace slow simulators and time-consuming experiments with still simple but highly adaptive mathematical models (emulators) that can run up to 1000 times faster. The use of emulators allows the adoption of complex multiobjective and global optimisation algorithms to search for an optimal solution. Emulators represent the best trade-off between accuracy and speed. One important issue is the proper adjustment (or setting) of these emulators. Together with other ideas in the project, this feature makes it possible to simultaneously find a design solution for both the product and the related processes which is optimised, fast, and robust

Different algorithms have been implemented for design centring, for example for the determination of the maximum-volume ellipsoid in a polytope and for the solution of problems involving both discrete and continuous variables. The last one is connected to a local search algorithm (sequential quadratic programming).

Risultati finali

Condividi questa pagina

Scarica