Artificial Neural Networks model for predicting wall temperature of supercritical boilers

• Published in: Applied thermal engineering Vol. 90; pp. 749 - 753
• Main Authors: Dhanuskodi, R., Kaliappan, R., Suresh, S., Anantharaman, N., Arunagiri, A., Krishnaiah, J.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.11.2015
• Subjects: ANN; Boiler design; Computational fluid dynamics; Fluid flow; Inner-wall temperature; Learning theory; Mathematical models; Neural networks; Supercritical; Turbulence; Turbulent flow; Wall temperature; ANN; Supercritical; Inner-wall temperature; Boiler design
• ISSN: 1359-4311
• DOI: 10.1016/j.applthermaleng.2015.07.036

Prediction of wall temperature for the range of operating conditions and selecting appropriate material for water-wall tubes, cooled by turbulent water/steam with drastic changes in property, is important in boiler design. An analytical route of predicting the wall temperature for such flow conditions is not reliable. Empirical correlations of non-dimensional numbers, based on experimental data, are used for predicting wall temperatures of turbulent flow with abrupt changes in fluid properties. BHEL has conducted many experiments with supercritical water/steam and developed Artificial Neural Network (ANN) based wall temperature prediction model. This model predicts wall temperature using the given inputs of fluid pressure, fluid temperature, product of mass flux and diameter, and heat flux. The model has prediction accuracy of 100% for the experimental data and 81.94% for the literature data at a deviation level of ±7 °C. This ANN model is useful for predicting wall temperatures of supercritical boilers operating in the tested range of parameters. [Display omitted] •Metal temperature is to be known at boiler design stage for material selection.•Experimental metal temperature data collected at supercritical water conditions.•ANN model developed using experimental data for metal temperature prediction.•100% agreement at ±7 °C deviation for experimental data.•97.22% agreement at ±10 °C deviation for literature data.