A new approach for reconstructing diesel engine cylinder pressure is presented. The technique is based on vibration measurements on the engine surface with subsequent reconstruction of the cylinder pressure by direct use of multivariate data analysis (MVDA). In order to investigate and evaluate the usefulness of the proposed technique, data from earlier experiments with four in-line, six cylinder, heavy-duty diesel engines have been used. One of the engines, running on ethanol, was tested according to a three factorial central composite face at different speeds and loads, as well as different blends of ethanol/ignition improver. The other engines, running on ordinary diesel, were tested with various loads and speeds. All of the measurements, i.e. cylinder pressure, sound pressure levels, vibration, and exhaust emissions, were performed simultaneously. The results demonstrated that MVDA models could accurately reconstruct cylinder pressures for all six cylinders in a diesel/ethanol engine. The differences between predicted and observed maximum cylinder pressure for 800 rpm were just 0-5%. The investigations also showed the potential of the method to estimate noise emissions and emission of NO x from the ethanol engine; a single partial least square (PLS) model could be used to predict noise and exhaust emissions at three different loads