Accueil Livres eBooks Revues spécialisées Références et comptes rendus Auteurs, éditeurs, examinateurs Index A-Z des produits Récompenses
Proceedings of an International Conference on Mitigation of Heat Exchanger Fouling and Its Economic and Environmental Implications

1-56700-172-6 (Imprimer)

Predicting Level Of Fouling Using Neural Network Approach

Anwar K. Sheikh
Department of Mechanical Engineering King Fahd University of Petroleum & Minerals Dhahran 31261, SAUDI ARABIA

M. Kamran Raza
Department of Mechanical Engineering King Fahd University of Petroleum & Minerals Dhahran 31261, SAUDI ARABIA

Syed M. Zubair
Mechanical Engineering Department, KFUPM Box # 1060, King Fahd University of Petroleum & Minerals, Dhahran 31261, Saudi Arabia

M. O. Budair
Department of Mechanical Engineering, King Fahd University of Petroleum and Minerals, Saudi Arabia


Physical fouling models provide tremendous insight about the parameters and variables, which play critical role in fouling growth. However due to time-dependent uncertainties of properties and operating conditions, as well as the complexities of some industrial fouling environment, it is quite difficult to make accurate prediction of fouling at a certain time. The uncertainties of fouling process can be incorporated in fouling prediction models by using a stochastic process formulation. These approaches [1-3], needs several replicate measurements of fouling at various time intervals to properly formulate a stochastic fouling process model. In this paper another promising approach using artificial intelligence will be illustrated to make accurate fouling predictions at a given time. The Neural Network has been widely used in various branches of engineering. We believe that this is a quite an attractive way to accurately predict the future fouling from the historical fouling data. This approach becomes extremely attractive in many complex situations where the physical models fail to provide the adequate results. The Neural Network can initially be trained on a limited data. As additional data and other relevant information is available a better and better prediction is possible. It is expected that a properly predicting Neural Network can be a powerful tool that a maintenance engineer can use to schedule cleaning of heat exchangers. In this regard, we will use fouling data from different sources to demonstrate the use of Neural Network technique to predict the time-dependent performance of a heat exchanger subject to fouling.