Prediction of Solid Holdup in a Gas–Solid Circulating Fluidized Bed Riser by Artificial Neural Networks
Hanbin Zhong, Zeneng Sun, Jesse Zhu, Chao Zhang
Abstract
The artificial neural network (ANN) method was applied to predict the solid holdup in a gas–solid circulating fluidized bed (CFB) riser. All the possible ANNs were first developed by looping the hidden neurons from the minimum (3) to the maximum (number of training data) and performing 500 independent runs for the same ANN structure. Then, an improved rule for finding the best ANN was proposed with the help of the expected range of the predicted solid holdup based on the existing data under training conditions. The accuracy of the prediction for test conditions was significantly enhanced by using the improved rule. The reproducibility and applicability of the proposed ANN development process were fully examined by repeating several times on the same sample and applying to different samples, respectively.