Prediction of combustion efficiency of chicken litter using an artificial neural network approach

This paper aimed to explore a feed-forward back-propagation artificial neural network (BPANN) approach to predict combustion efficiency of chicken litter in a swirling fluidized bed combustor. A series of experiments were conducted based on the statistics-based design of experiment method. The data for combustion efficiencies under various operational conditions were obtained to train artificial neural network. The operational conditions were adjusted through moisture content in waste, excess air, litter ratio, secondary air and its injection height. A BPANN was constructed and trained with two training algorithms: Levenberg–Marquardt and gradient descent. The response surface of combustion efficiency along with the five parameters was accordingly predicted. The results showed that for the same mean squared error (0.2204) the Levenberg–Marquardt training algorithm is much faster than the gradient descent. The best architecture of neural network was found as 5 + 16 + 1. Also, the predicted response surface clearly showed how combustion efficiency changes along with the operational parameters. Moreover, the relative high combustion efficiency (over 84%) was found within the ranges: moisture content as 11–14%, litter ratio as 0.05–0.1, excess air as 0.22–0.45, secondary air as 0.18–0.27. A validation experiment under these conditions showed that the artificial neural network approach provides an easy and reliable prediction for combustion efficiency.