Optimization of operating conditions for compressor performance by means of neural network inverse

A way to optimize the parameters (i.e. operating conditions), related to compressor performance, based on artificial neural network and the Nelder–Mead simplex optimization method is proposed. It inverts the neural network to find the optimum parameter value under given conditions (artificial neural network inverse, ANNi). In order to do so, first an artificial neural network (ANN) was developed to predict: compressor pressure ratio, isentropic compressor efficiency, corrected speed, and finally corrected air mass flow rate. Input variables for this ANN include: ambient pressure, ambient temperature, wet bulb temperature, cooler temperature drop, filter pressure drop, outlet compressor temperature, outlet compressor pressure, gas turbine net power, exhaust gas temperature, and finally fuel flow mass rate. For the network, a feed-forward with one hidden layer, a Levenberg–Marquardt learning algorithm, a hyperbolic tangent sigmoid transfer-function and a linear transfer-function were used. The best fitting with the training database was obtained with 12 neurons in the hidden layer. For the validation of present database, simulation and experimental database were in good agreement (R2>0.99)(R2>0.99). Thus, the obtained ANN model can be used to predict the operating conditions when input parameters are well-known. Second, results from the ANNi that was developed also show good agreement with experimental and target data (error <0.1%), in this case, cooler temperature was found for a required efficiency. Therefore, the proposed methodology of ANNi can be applied to optimize the performance of the compressor with an elapsed time minor to 0.5 s.