Neural Modeling of the Energy Efficiency Factor for Recuperators (Heat Exchangers) Using Impinging Jets for the Metallurgical Productions and Machine Building

The given work contains a complete application analysis for neural networks of the common logarithm of the energy factor for heat exchangers (recuperators) to heat the air from the heat exchange surface onto which the systems of impact plane-parallel jets are delivered in heating and thermal furnaces in machine building and at the metallurgical plants. The following input signals have been provided: the width of the slot nozzle, the jet axial length, the nondimensional length from the heat transfer surface to the nozzle cut (related to the width of the slot nozzle); the specific (per unit area of the heat transfer) consumption of air. The matching of the target records of the common logarithm of the energy efficiency factor with the output values was carried out (obtained using the artificial neural network in case of the network learning applying a quasi-Newton technique, BFGS, and the method of random increments) for the network structure 4-9-9-9-1. The analysis of regression for the obtained data and target outputs was implemented, and the correlation factors were given. The optimal results with the lesser values of Mean Squared Error and the closest to one of the values of the correlation coefficients were obtained when using the following algorithms: used as the learning function of adaption of the training function for the descent of a gradient with the momentum, as the transfer function—the hyperbolic tangent and the network learning using the quasi-Newton technique, BFGS. The results obtained can be applied in order to optimize or to develop the calculation techniques for the new highly efficient heat exchange devices exploiting the systems of plane-parallel impingement jets as well as for a wider use of the impinging jets technology in various industrial branches such as design and manufacturing for different new technologies including mechanical facilities.