Intelligent Systems*

Professor Mihir Sen
Department of Aerospace and Mechanical Engineering
University of Notre Dame
Notre Dame, IN 46556

*Sponsored by D.K. Dorini (BRDG-TNDR), Organization of American States, CONACyT-Fulbright.


Overview


Intelligent systems have been used in heat transfer applications in several ways. Though the examples given below mainly involve heat transfer rates in heat exchangers, similar techniques would apply to other complex problems in thermal engineering such as to the prediction of pressure drops, etc. Though a brief description is given of each, the details are left to the publications. Given the form of a heat transfer correlation, the genetic algorithm provides a way to determine the globally optimal regression values of the coefficients. This is under an ideal heat exchanger assumption from which there can be deviations. A better alternative to using correlations is to use artificial neural networks to learn from experimental data and then to use the network to predict the performance of the heat exchanger for other operating conditions. The error is much less than for conventional correlations. Neural networks and fuzzy logic modeling can be used to predict statistical quantities like mean and rms velocities in turbulent flows. Heat transfer surfaces are susceptible to fouling and change in thermal characteristics over time. Artificial neural networks are a good way to learn, to adapt to changing circumstances, and to be able to control thermal systems. Techniques have been developed to obtain a network that is dynamically stable and that enables efficient temperature control.

Experimental facilities

The experiments have been carried out in the Hydronics Laboratory in the Fitzpatrick Hall of Enginering. The laboratory facilities include the following.

Research

Theses and dissertations

Teaching