Input conversion network

The key parts of the positioner/actuator system, shown in Fig. 8-74 7, are (1) an input-conversion network, (2) a stem-position feedback network, (3) a summing junction, (4) an amplifier network, and (5) an actuator. 

The stem position feedback network converts stem travel to a useful form for the summer. This block includes the feedback linkage which varies with actuator type. Depending on positioner design, tKe stem position feedback network can provide span and zero and characterization functions similar to that described for the input conversion block. 

In another example, a neural network was applied successfully to the prediction of conversion and yields in the decomposition of NO into N2 and O2 over Cu/ ZSM-5 catalysts [42]. This shows that artificial neural networks arc able to describe a complex catalytic system quite well if an appropriate numerical representation is found for the input and output data. The problem in their application may be the availability of a representative set of experimental data for learning, as well as the interpretation of the weights obtained for neuron interconnections which do not enable direct derivation of guidelines for the optimization of a catalytic system. 

Using artificial neural networks (ANN) the reaction system, including intrinsic reaction kinetics but also internal mass transfer resistances, is considered as a black-box and only input-output signals are analysed. With this approach the conversion rate of the i-th reactant into the j-th species can be expressed in a general form as a complex function, being a mathematical superposition of all above mentioned functional dependencies. This function includes also a contribution of the internal diffusion resistances. So each of the rate equations of Eq. 5 can be described with the following function based on the vanables which uniquely define the state of the system ... 

---