Flow estimation using genetic algorithm and neural network

The paper presents flow estimation method of parallel driven pumps with bended pipes near pumps´ output parts. Flow can not be measured near bended pipe region due to turbulence of fluid. So, it needs to find other methods to estimate the flow. In this paper, we propose flow estimation method using genetic algorithm (GA) and neural network (NN). Parallel driven pumps are modeled using NN and the weights of NN are learned from GA through fitness evaluation. Fitness functions are defined by average value of errors between measured flow values of main pipe and estimated values of it. Max/min value of each pump´s flow is constrained in order to reduce search space of GA and to raise precision of estimation. The effectiveness of proposed algorithm is proven through experiments.