This paper proposes a real-time combustion control algorithm using reconstructed in-cylinder pressure traces by principal component analysis (PCA). The PCA method reconstructs the in-cylinder pressure traces using the principal components of the in-cylinder pressure traces. It was shown that using only five principal components, we were able to reconstruct the in-cylinder pressure traces within 1% root mean squared percent error. Furthermore, the reconstructed in-cylinder pressure traces were validated to effectively reduce the cycle-to-cycle variations caused by the noise signals. As a result, the standard deviation of MFB50 which was calculated from the reconstructed in-cylinder pressure was reduced by 45%. Furthermore, this combustion parameter was applied to a real-time combustion control. Since variations of the control variables for the real-time combustion control were reduced, the control performances were enhanced.