Principal component analysis (PCA) is widely used in chemical process monitoring. It selects the first several principal components (PCs) with the most variances information of normal observations for process monitoring. However, PCA may ignore fault information contained in the subspace spanned by the rejected PCs. In this paper, we propose a novel algorithm called sparse principal component selection (SPCS). SPCS can be formulated as a just-in-time reorganized PCA algorithm that constructs an elastic net regression between all PCs and each sample. SPCS selects PCs according to the non-zero regression coefficients which indicate the compact expression of the samples. This expression is naturally discriminative: amongst all subset of PCs, SPCS selects the PCs that most compactly expresse the samples and reject all other possible but less compact expressions. The case studies on the Tennessee Eastman process demonstrate the effectiveness of SPCS on process monitoring. The performance of SPCS is significantly better than other PCA based algorithms.