Robust calibration of computer models of engineering networks with limited data

Many practical engineering problems are characterised by the lack of sufficient accurate data to permit a conventional computer-based simulation, analysis or optimization to be performed. This paper describes a new method for logically inferring values for missing data items in engineering network problems. The method is based upon entropy maximization and permits the calibration of engineering networks to be performed robustly and with minimum bias. The theoretical basis of the method is outlined and two water supply network examples are used to demonstrate that the method can be successfully implemented. The performance of the method on the two examples is discussed.