This paper describes a predictive model of NO X and consumption oriented to control and optimisation applications in DI Diesel engines. The model, which is based on the Response Surface Methodology (RSM), is focused on the prediction of NO X emissions and brake specific fuel consumption (output parameters) following a two step process: first, the relationship between engine inputs and combustion parameters is determined and, secondly, engine outputs are predicted from the combustion parameters, taking into account the mechanical losses. The inputs are engine variables related to the intake charge conditions (exhaust gas recirculation rate, intake pressure and temperature), injection settings (start of main and pilot injections, and injection pressure), and some combustion parameters that allow to characterise the in-cylinder gas processes (peak pressure, indicated mean effective pressure and burned angles). Splitting the model into two parts allows using either experimental or modelled combustion parameters, thus enhancing the model flexibility. The proposed model is finally used for multi-objective optimisation of engine operation.