Currently, diesel engine-out exhaust NOx emission level prediction is a major challenge for complying with the stricter emission legislation and for control purpose of the after-treatment system. Most of the NOx prediction research is based on the Zeldovich thermal mechanism, which is reasonable from the physical point of view and for its simplicity. Nevertheless, there are some predictable range limitations, such as low temperature with high EGR rate operating conditions or high temperature with low EGR rates.In the present paper, 3 additional considerations, pilot burned gas mixing before the main injection; major NO formation area; concentration correction, were applied to the previously developed real-time NO estimation model based on in-cylinder pressure and data available from ECU. The model improvement was verified on a 1.6 liter EURO5 diesel engine in both steady and transient operation. These considerations expand the predictable NO concentration range from 15 ppm up to 1461 ppm in the steady state conditions with an R2-value of near 0.98, although EGR rate, main injection timing, boost pressure and rail pressure varied. In addition, the NOx prediction results during the WLTP cycle matched the portable NOx sensor and HORIBA MEXA 7100.This virtual NOx sensor concept might be useful as an engine emission calibration tool, for real-time control of exhaust NOx emission levels, and as an alerting system for the on-board NOx sensor malfunction.