Inferential Statistics Using Partial Least Square Regression (PLSR) for Chemical Processes. Case Study: A Fixed Bed Catalytic Reactor

The complexity of processes in the modern industry (many kinetics and fluid dynamic variables) makes to think in few data due to high costs for experimental data collection. Therefore, it is important to consider statistical models, such as the partial least square regression (PLSR) method that provide an adequate solution to the multicollinearity in regression models, due to the minimum requirements in terms of measurement scales, sample size, and residual distribution. This work has shown for first time the application of the PLSR method to a fixed bed catalytic reactor for cracking of t-pentene-2 to obtain light olefins. Results show a correlation of R²_predictive > 0.85, which was validated with other experimental data of the same chemical process indicating the robustness of the method as a powerful statistical tool to infer the behavior in chemical process. The importance of this research lies to present the use of PLSR in engineering applications as such rating and design of chemical processes as a novel alternative method to predictive behaviors with few experimental data tests.