This paper presents a comprehensive review of the historical development, the current state of the art, and prospects of data-driven approaches for industrial process monitoring. The subject covers a vast and diverse range of works, which are compiled and critically evaluated based on the different perspectives they provide. Data-driven modeling techniques are surveyed and categorized into two main groups: multivariate statistics and machine learning. Representative models, namely principal component analysis, partial least squares and artificial neural networks, are detailed in a didactic manner. Topics not typically covered by other reviews, such as process data exploration and treatment, software and benchmarks availability, and real-world industrial implementations, are thoroughly analyzed. Finally, future research perspectives are discussed, covering aspects related to system performance, the significance and usefulness of the approaches, and the development environment. This work aims to be a reference for practitioners and researchers navigating the extensive literature on data-driven industrial process monitoring.

The proper design of RTO systems’ structure and critical diagnosis tools is neglected in commercial RTO software and poorly discussed in the literature. In a previous article, Quelhas et al. (Can J Chem Eng., 2013, 91, 652–668) have reviewed the concepts behind the two-step RTO approach and discussed the vulnerabilities of intuitive, experience-based RTO design choices. This work evaluates and analyzes the performance of industrial RTO implementations in the face of real settings regarding the choice of steady-state detection methods and parameters, the choice of adjustable model parameters and selected variables in the model adaptation problem, the convergence determination of optimization techniques, among other aspects, in the presence of real noisy data. Results clearly show the importance of a robust and careful consideration of all aspects of a two-step RTO structure, as well as of the performance evaluation, in order to have a real and undoubted improvement of process operation.

Data reconciliation is a model-based technique that reduces measurement errors by making use of redundancies in process data. It is largely applied in modern process industries, being commercially available in software tools. Based on industrial applications reported in the literature, we have identified and tested different configuration settings providing a numerical assessment on the performance of several important aspects involved in the solution of nonlinear steady-state data reconciliation that are generally overlooked. The discussed items are comprised of problem formulation, regarding the presence of estimated parameters in the objective function; solution approach when applying nonlinear programming solvers; methods for estimating objective function gradients; initial guess; and optimization algorithm. The study is based on simulations of a rigorous and validated model of a real offshore oil production system. The assessment includes evaluations of solution robustness, constraint violation at convergence, and computational cost. In addition, we propose the use of a global test to detect inconsistencies in the formulation and in the solution of the problem. Results show that different settings have a great impact on the performance of reconciliation procedures, often leading to local solutions. The question of how to satisfactorily solve the data reconciliation problem is discussed so as to obtain improved estimates.

Digital twins are rigorous mathematical models that can be used to represent the operation of real systems. This connection allows for deeper understanding of the actual states of the analyzed system through estimation of variables that are difficult to measure otherwise. In this context, the present manuscript describes the successful implementation of a digital twin to represent a four-stage multi-effect evaporation train from an industrial sugar-cane processing unit. Particularly, the complex phenomenological effects, including the coupling between thermodynamic and fluid dynamic effects, and the low level of instrumentation in the plant constitute major challenges for adequate process operation. For this reason, dynamic mass and energy balances were developed, implemented and validated with actual industrial data, in order to provide process information for decision-making in real time. For example, the digital twin was able to indicate failure of process sensors and to provide estimates for the affected variables in real time, improving the robustness of the operation and constituting an important tool for process monitoring.

Predictive analytics is usually cited as one of the most important pillars of the digital transformation. For the oil industry, specifically, it is a common belief that issues like integrity and maintenance could benefit from predictive analytics. This paper presents the development and the application of a process-monitoring tool in a real process facility. The PMA (Predictive Maintenance Application) system is a data-driven application that uses a multivariate analysis in order to predict the system behavior. Results show that the use of a multivariate approach for process monitoring could not only detect an early failure at a metering system days before the operation crew, but could also successfully identify, among hundreds of variables, the root cause of the abnormal situation. By applying such an approach, a better performance of the monitored equipment is expected, decreasing its downtime.