A comparison of fuzzy expert systems, neural networks and neuro-fuzzy approaches. Controlling energy and material flows

Abstract

In industrial production processes, materials and different forms of energy are provided, transformed respectively converted, stored and transported. With this process joint products in different states of aggregation are emitted. Environmental impacts can be identified at any stage of the energy and material flow process. Due to the fact that production units and processes are interconnected with energy and material flows, it is of special interest to develop production control mechanisms which control the energy and material streams in a way that utilizes available resources most efficiently and reduces emissions and by-products caused by the production process. These production control strategies have to consider variations in the input and output flows of succeeding and preceding production units. The development of production control strategies depends especially on the structure of integrated production systems. If it is possible to influence the energy and material flows by the selection of special production processes and an adequate allocation of jobs and aggregates, the construction of production control strategies can be reduced to a combined scheduling and technology selection problem. Methodical production control strategies can be based on optimal algorithms (e.g. dynamic programming) heuristics (e.g. rule-based approaches) and methods of machine learning (e.g. neural networks). Due to the complexity of real production systems, it is advisable to use rule-based approaches or neural networks depending on the structure of the available production knowledge.