Probabilistic Graphical Models for Data Reconciliation and Causal Inference in Process Data Analytics

  • Author / Creator
    Arun Senthil Sundaramoorthy
  • Data reconciliation methods play an important role in minimizing the measurement error and gross error that are present in the process data with respect to the process model. On the other hand, causal analysis helps in determining the relationship among the process variables from the data. It is important to note that these aforementioned methods help in understanding the process data better, provide valuable insights of the process and contribute to improving the decision making strategies in the day to day process operation. In this thesis, we develop approaches based on probabilistic graphical models to address the problems in data reconciliation and causal inference. In the first part of the thesis, the problem of data reconciliation with state uncertainties is addressed in the framework of probabilistic graphical models. In many existing formulations for data reconciliation, process models are assumed to be error free. However, while operating the process in real time, process models can suffer from model inaccuracies, leading to uncertainties in states. This work introduces a new method for data reconciliation developed in the framework of Bayesian network, accounting for the state uncertainties. A novel method to construct acyclic Bayesian networks for process networks with recycle streams is proposed in this work. This method is also extended for data reconciliation of partially measured systems. The solution is obtained by utilising a Bayesian network model translated from the process model and using statistical inference techniques to estimate the reconciled values of the states. The efficacy of the proposed data reconciliation schemes is demonstrated on three case studies namely Simple Flow Network, Mineral Processing Unit (without Recycle) and Mineral Beneficiation Process (with Recycle). In the second part of the thesis, novel methods are proposed for causal network modelling. First, a framework for causality analysis based abnormal event prediction is proposed. Also, a methodology to construct causal network of the process systems using finite impulse response model with sparsity constraints and process knowledge is developed. Using causal network reconstructed from data and the process knowledge, we develop and test process monitoring hypotheses. Efficacy of the proposed approach is illustrated using a real industrial process case study of flooding and weeping in a distillation column. Further, we discuss the outcomes and the findings from the field implementation of process monitoring framework. Further, a methodology for causal analysis is developed that makes use of only the process data without any information about the process knowledge. In this method, the sparse inverse covariance estimation is coupled with dynamic likelihood score, and a two-step approach is proposed to address the problem of causal analysis. The estimation of sparse inverse covariance matrix for undirected sparse network reconstruction is performed with $L_0$ norm constraint in the framework of greedy sparse simplex (GSS) algorithm. Further, the GSS algorithm is suitably modified to incorporate the additional constraint of positive semi-definiteness of the inverse covariance matrix. To determine the causal direction among the variables, dynamic likelihood score is computed for the associated variables in the reconstructed network in the second step. The efficacy of the proposed approach for causal analysis is illustrated using a numerical example and an industrial application on prediction of flooding and weeping in a deethanizer column associated with a fluid catalytic cracking unit.

  • Subjects / Keywords
  • Graduation date
    Spring 2021
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.