Identification and control designs are considered using neural networks for a class of nonlinear partially known dynamic systems. Real-time implementation of two designs, a neural identifier and a proposed neural controller, using an experimental system, comparisons with two other neural networks as well as conventional schemes, and an implementation architecture are reported. The proposed control design facilitates incorporation of available knowledge about the structure of the system. The study also illustrates the inherent capability of neural networks to handle nonlinearities and perform control effectively for a real world system, based on minimal system information.

No abstract

The dual goals of this study are: (1) to develop an empirically valid neural model of U.S. factories in a range of industries producing discrete products, and (2) to use the model to test the effect of changes in product line width on plant performance variables. Accordingly, a neural factory was developed using 59 input and 5 output/performance variables, and was trained using field data collected from 385 U.S. manufacturing plants. The model was validated using a holdout sample before conducting sensitivity tests. The study demonstrates that, through the use of parametric sensitivity analysis, the neural factory could be used to investigate the relationship between inputs and performance of a factory. While the focused factory principle would favor a smaller product line, economies of scope theory would favor a larger product line for the good of the factory; this implies a rather complex relationship between product line width (PLW) and plant performance. The neural factory was used to study the sensitivity of output/performance variables when product line width was varied over a range extending from 10% to 200% of the average values. The sensitivity analysis of the neural factory shows that, as the product line increases, it (1) does not affect cost-of-goods-sold (COGS), (2) decreases return on investment, (3) has a negative effect on the top management's perception of manufacturing performance, (4) increases inventory turns, and (5) increases sales per employee. The explanations for these findings show how complex and intertwined the relationships between PLW and performance variables are. They enhance our understanding of PLW and provide some new directions for future empirical research.neural factory, factory model, manufacturing plant model, neural network, neural model of manufacturing, factory focus, product line width

No abstract