Simulation and control of heat exchangers using artificial neural networks.

摘要

The design of thermal systems usually requires the prediction of heat transfer rates of heat exchangers under prescribed operating conditions. Due to the complexity of these thermal components, conventional steady-state modeling approaches, such as correlations, provide predictions with large uncertainties. These are not only due to experimental errors but also to the information compression process in which several assumptions are used. For control purposes, furthermore, dynamic simulations are needed for which only a limited number of models are available. We apply artificial neural networks (ANNs) to the simulation of the steady and dynamic behaviors of heat exchangers, as well as to the control of fluid temperatures. The experiments were carried out in a heat exchanger test facility. The ANN predictions are obtained using information about the flow rates and inlet temperatures of both fluids in the heat exchanger. Numerical tests show the feasibility of the method and experimental comparison with conventional correlations prove the ANN to be more accurate.; Dynamic prediction is also addressed with analytic and experimental evidence of excellent predicting characteristics of the ANNs. Using dynamic modeling, ANNs are used in conjunction with internal model control to perform non-adaptive and adaptive control of the air temperature leaving a single-row water-to-air fin-tube heat exchanger. Stability constraints are included in the training of the ANNs. The closed-loop system is considered as a nonlinear iterative map and its stability is analyzed numerically and verified experimentally. Reduction in the energy consumption is added as one of the tasks of the neurocontroller.; Finally, the delay effects involved in the thermal system due to sensor location are analyzed. Analytical and experimental comparisons with conventional on-off control are performed and model predictive control using ANNs to simulate the physical plant is used to improve the performance of the conventional on-off control scheme in the presence of delay. It is shown how the system remains within the dead band of the on-off control system with the use of an ANN predictive model.