Taking the particulate matter emissions of turbocharged and intercooled heavy-duty diesel engine at four different altitudes as the research object, the particle size concentration under real driving condition was simulated and analyzed by the combination of principal component analysis and neural network. The results show that the first 10 principal components of cylinder pressure can represent 94% of incylinder combustion characteristics at different altitudes. The nuclear modal particles are little and the accumulated modal particles are many, especially for those with particle size of 57 165 nm. In addition, the model can effectively predict the particulate concentration within the range of 7 990 nm at four altitude areas with the relative error reduced by 6.44% and the prediction accuracy of 91.37%, 92.97%, 91.23% and 91.99% respectively compared with the traditional model. The research provides support for monitoring and controlling the pollutant emissions in plateau area.