1- OPA

The Orthogonal Projection Approach (OPA) is a self-modelling curve resolution method. OPA could be employed during process monitoring to obtain concentration profiles and/or pure spectra of a mixture. This method works in two steps:

1- In the first step of this method, the right number of components presents in the mixture has to be selected according to a dissimilarity criterion. This step is completely interactive and the user has to select as many components as he thinks necessary. This selection should be stopped when random dissimilarity profiles are obtained (see theory in the reference for more details).

2- After this selection, the second step is Multivariate Curve Resolution using Alternating Least Squares (MCR-ALS). It is applied to resolve the data matrix X into the pure component spectra and their related individual concentration profiles.

For a scientific overview of the method, click here.

On a 2-way matrix (spectra by variables for instance), OPA could be performed in the two directions.



Dissimilarity plots are then available to select "pure variables" (pv) or "pure spectra" (ps)



On these dissimilarity plots, one can decide to select pv / ps or to stop the selection to go to the resolution step of OPA (MCR-ALS).
To select one pure variable or one pure spectrum, you click on "Select the maximum" and you get the next dissimilarity plot. If you want to come back to the previous one and to cancel the selection, go to "Reset" menu (at the top).

Before performing the resolution, you can set some parameters and also some constraints



After this window, the iterative resolution procedure starts...