Multivariate methods such as partial least squares (PLS), interval PLS, and other variants are often the default option for prediction of lubricant properties based on FTIR spectra. However, advanced analytical methodologies also available that have not been properly tested comparatively assessed so far. The present work focuses comparison predictive ability four classes methods: regression with variable selection, penalized regression, latent tree-based ensemble methods. A data set 62...

Wastewater treatment plants (WWTPs) are complex systems presenting stochastic, non-linear, and non-stationary behavior, which makes their operational management very challenging. In this context, data collected from distributed sources across the plant play a central role in optimized operation control of WWTPs. However, even when available, use is far trivial due to coexistence asynchronous measurements, with different granularity, measurements quality (precision, accuracy), multimodal...

Abstract The new EU Regulation urges shipping operators to set up systems for the monitoring, reporting, and verification of CO 2 emissions. Indeed, monitoring data acquisition installed on modern ships have brought a navigation overload that needs be correctly handled make proper decisions about their operation. However, in today's market, there is no standard solution or method available can robustly adopted real environments industry. In view novel attempts solving this issue proposed by...

We present a data analytics framework for offline analysis of batch processes. The provides unified setting implementing several variants feature oriented proposed in the literature, including new methodology based on process variables' profiles presented this article. It also integrates generation and analysis, order to speed up exploration cycle, which is especially relevant complex FOBA (Feature Oriented Batch Analytics platform) described detail applied case studies regarding different...

Obtaining and handling measurement uncertainty information is still a challenge in the chemical processing industries (CPI). From our experience, among variables most affected by uncertainty, one typically finds process outputs, comprising concentrations (main product subproducts, reactants, etc.), measurements of quality properties (mechanical, chemical, or other relevant about end use product. With increasing flexibility units, these quantities can easily span different orders magnitude...

A significant number of batch process monitoring methods have been proposed since the first groundbreaking approaches were published in literature, two decades ago. The proper assessment all alternatives currently available requires a rigorous and robust framework, order to assist practitioners their difficult task selecting most adequate approach for particular situations they face definition optional aspects required, such as type preprocessing, infilling, alignment, etc. However,...

Abstract In this article, multiple reinforcement learning (RL) methods such as value‐based, policy‐based, and actor‐critic algorithms are investigated for typical control tasks found in the chemical industries. Through a critical assessment of these novel techniques, their main advantages highlighted, but also challenges that still need to be resolved discussed. Two batch used benchmarks, namely, production maximization, setpoint control. Using testing environments, direct comparison...

Signal denoising is a pervasive operation in most online applications, such as engineering process control and optimization, strongly affecting the outcome of these higher‐level tasks impacting overall variability exhibited by processes products. Therefore, it plays fundamental role improving capability, which is, however, often overlooked. In this work, we compare performance different types currently available filters using variety test signals that represent diversity situations likely to...

Equipment degradation is ubiquitous in the Chemical Process Industry (CPI), causing significant losses efficiency, controllability, and plant economy, as well an increased environmental fingerprint additional operational safety risks. The case of fouling heat exchangers, particular, well-known pervasive but still hard to cope with, given complexity underlying mechanisms difficulty assessing its extension real-time. This problem becomes even more complex batch processes producing different...

Abstract Shipping companies are forced by the current EU regulation to set up a system for monitoring, reporting, and verification of harmful emissions from their fleet. In this regulatory background, data collected onboard sensors can be utilized assess ship's operating conditions quantify its CO 2 emission levels. The standard approach analyzing such sets is based on summarizing measurements obtained during given voyage average value. However, compression step may lead significant...

In multisensor fusion, several sources of information are combined in order to increase the estimation quality for quantity interest. This activity finds many applications from tactical missile defense self-driving cars and variables difficult measure such as analyte concentrations chemical processes. industrial applications, it is common employ laboratory analysis that provides more accurate measurements but usually at slower rates, with significant delays requiring involvement highly...

This paper presents an extended comparison study between 16 different linear and non-linear regression methods to predict the sugar, pH, anthocyanin contents of grapes through hyperspectral imaging (HIS). Despite numerous studies on this subject that can be found in literature, they often rely application one or a very limited set predictive methods. The literature multivariate is quite extensive, so analytical domain explored too narrow guarantee best solution has been found. Therefore, we...

Since the first batch process monitoring approaches were published in literature approximately 20 years ago, a significant number of extensions and new contributions have been proposed. They enrich toolkit solutions made available to practitioners, who face today daunting task finding best tool among all them, as well defining tuning associated configuration options. This could be greatly facilitated by availability sound comparison studies, with rigorous unambiguous metrics language....

In this paper, we implement a framework which combines Reinforcement Learning (RL) based reaction optimization with first principle model and plant historical data of the system. Here employ Long-Short-Term-Memory (LSTM) network for surrogate modeling, Proximal Policy Optimization (PPO) algorithm fed-batch optimization. The proposed simulation real an accurate computationally efficient simulation. Based on model, RL result suggests maintaining increased temperature setpoint high reactant...

The operational management of wastewater treatment plants (WWTP) is a complex activity due to the biological phenomena' intricate nature. This complexity hinders adoption first principles approaches, which lack necessary accuracy be adopted in practice. Data-driven methodologies also face significant challenges processing different information sources available. In this work, we present data-driven and model-agnostic data-fusion framework estimate concentration level toxin effluent, using...

Reinforcement learning is a branch of machine learning, where an agent gradually learns control policy via combination exploration and interactions with system. Recent successes model-free reinforcement (RL) has attracted tremendous attention from the process community. For instance, RL been successfully applied in very complex tasks (e.g., games such as chess or Go that contain large state spaces) shown to be robust uncertainties. These findings indicate there significant potential leverage...