Artificial neural network models for knowledge representation in chemical engineering

Abstract Artificial neural network models exhibit the rule-following behavior of knowledge-based expert systems without containing any explicit representations of the rules. Instead of using explicit rules, the neural network model relies on a number of very simple neuron-like processing elements locally interacting through a set of unidirectional weighted connections. Knowledge is internally represented by the values of the weights and the topology of the connections. Learning involves modifying the connection weights. These networks can learn and adapt themselves to inputs from the actual processes allowing representation of complex engineering systems, which are difficult to model either with traditional model-based engineering or knowledge-based expert systems. This paper describes the characteristics of neural networks desirable for knowledge representation in chemical engineering processes. It explains the characteristics of neural networks focusing on the nodes, the connections and their learning capabilities, and describes a neural network design and simulation environment that can be used for experimentation. To illustrate the concepts, we demonstrate via a simple chemical engineering process how an artificial neural network can learn and discriminate successfully among faults.