International Journal of Engineering, Science and Technology

The AJOL site is currently undergoing a major upgrade, and there will temporarily be some restrictions to the available functionality.
-- Users will not be able to register or log in during this period.
-- Full text (PDF) downloads of Open Access journal articles will be available as always.
-- Full text (PDF) downloads of subscription based journal articles will NOT be available
We apologise for any inconvenience caused. Please check back soon, as we will revert to usual policy as soon as possible.

Optimization of multi-response dynamic systems integrating multiple regression and Taguchi’s dynamic signal-to-noise ratio concept

Surajit Pal, Susanta Kumar Gauri


The principal difference between a dynamic and a static system is that the former includes a signal factor for expressing the intended output while the later does not. Assuming a linear association exists between the response and signal variables, Taguchi offered a two-stage route for optimizing a dynamic system: maximize the dynamic signal-to noise ratio (DSN) and then, change the gradient to the desired gradient by a suitable modification parameter. Some researchers have indicated limitations to Taguchi’s DSN analysis, and advocated alternative approaches for optimization of a dynamic system. However, the Taguchi method as well as these alternative approaches is useful for optimizing a single-response dynamic system only. In realism, the majority of the contemporary manufacturing practices encompass numerous response variables as well as industries demand for developing procedures for optimizing multi-response dynamic system. This paper proposes a novel procedure that integrates multiple regression (MR) technique and Taguchi’s DSN concept to optimize the multi-response dynamic system. In this method, appropriate multiple regression equations according to a chosen model for dynamic system are fitted first based on the observed experimental data and then DSN (called MRDSN) for different response variables are computed using the MRbased predicted values. Finally, weighted MRDSN is considered as the objective function for the optimization. The proposed procedure is investigated with respect to three modelling approaches for the dynamic systems. The results of analysis reveal that the proposed procedure with response modelling approach results in the best optimization performance. It also results in better optimization performance than back-propagation neural network-based approach and data mining-based approach reported by the past researchers.

Keywords: multiple responses, multiple regression, weighted dynamic signal-to-noise ratio, performance measure modelling, response function modelling, response modelling, optimization

AJOL African Journals Online