Spectroscopy is a useful tool for analyzing chemical information in batch processes. However, conventional spectroscopic techniques call lead to complex model structures and difficult-to-interpret results because of the direct use of redundant wavelength variables. To solve this problem, the ICA technique has been used first to reveal the underlying independent sources from the observed mixtures and their mixing coefficients. This analysis associates the constituent species with their respective effects on the mixture spectra, which could make more sense for spectroscopy. In the following article, the use of mixing coefficients instead of redundant wavelength variables can provide a convenient modeling platform from which the phase-specific characteristics of chemical reactions are readily identified. Consequently, a phase-based joint modeling framework is formulated For process monitoring by combining different latent-variable-(LV-) based algorithms. It decomposes different types of systematic variations in spectral measurements (X) and then sets Lip different monitoring systems for them. Monitoring different parts of X can provide abundant process information about the Status of the operation from a comprehensive viewpoint, thus benefitting process understanding and Fault detection. The effectiveness of this approach is demonstrated by application to two case studies.