Imperial College London

Dr. Salvador García Muñoz

Faculty of EngineeringDepartment of Chemical Engineering

Visiting Professor







ACE ExtensionSouth Kensington Campus





Multivariate Latent Variable Methods (LVM)

  • Theory, algorithms and parameter estimation solutions for reduced rank projection and regression methods (PLS,PCA)
  • Application of multi-block and multi-path methods to complex data structures.
  • Optimization solutions with embedded  latent variable models for assisted process operations.
  • Monitoring, troubleshooting, fault-detection control and optimization of process systems with LVM.

algorithms and methods for process analytical technology

  • Lean development and low cost of ownership methods for PAT (EIOT).
  • Self-sustaining methodologies (structured adaptive techniques).
  • Data fusion.
  • Integration of PAT metrics into state estimation, reconciliation and non-linear control solutions

statistics and fundamental modeling

  • Model Based Experimental Design for non-linear differential algebraic systems. i) Model Discrimination, ii) Model Parametrization, iii) Model Exploration.
  • Uncertainty handling and analysis for indirect measurements (PAT) incorporated into fundamental models.

hybrid modeling systems

  • Process monitoring, state-estimation and fault detection using fundamentally derived  and empirically derived models.


  • Application of optimization methods to all of the above.

pyPhi - A Python package for Multivariate Analysis

show research

pyphi is a python based package for multivariate analysis.

Version 1.0 includes: Principal Components Analysis, Projection to Latent Structures, LWPLS, Savitzy-Golay derivative transform, Standard Normal Variate transform. PCA and PLS routines support to missing data.

pyphi_plots is a package with a variety of plotting tools for models created with pyphi


Dependencies: numpy, scipy, pandas, datetime, bokeh , matplotlib

The packages can be downloaded from: