This study presents practical details and results from implementing model-based predictive controllers for artificial lift systems in oil production, which utilise electrical submersible pumps (ESPs) in legacy systems. The proposed methodology integrates the existing instrumentation architecture with new control systems and techniques. This work implements two control strategies: a standard MPC and an MPC with input targets. Both were executed in real time, encapsulated in a C++ executable, and deployed within Petrobras’s supervisory and control system. The results include an analysis of the instrumentation system, a discussion of operation under unmeasured disturbances and constraints, and a comparison of the controllers. The findings are extensive and indicate that both controllers stabilise the system, ensure constraint satisfaction, and appropriately compensate disturbances.

The complexity of multiphase flow systems poses significant engineering challenges. For Core Annular Flow (CAF) systems, where a viscous core is enveloped by a less viscous fluid, accurate flow pattern identification is crucial for optimizing efficiency. This research pioneers an integrated approach using Computational Fluid Dynamics (CFD) and fuzzy clustering to identify and classify CAF patterns. A novel methodology was developed: numerical simulations provide data on velocity and phase distributions, which are then processed using principal component analysis (PCA) and a fuzzy logic-based clustering algorithm specifically designed to resolve transitional flow states. Our findings confirm the methodology’s capacity to effectively differentiate various flow regimes and, importantly, to characterize the continuous transitions between them. This work delivers a robust and adaptable framework for improved understanding and optimization of CAF in industrial contexts.

This article describes the construction and validation of a system applied to measuring the temperature of shelves and load cells in a freeze dryer using thermistors as sensors. Conditioning and analog-to-digital conversion circuits were built, two linearization techniques were applied to the thermistor response, and calibration was performed. A freeze-drying process was carried out to validate the system using sliced yam samples. The results indicated a maximum linearization error of 0.4 degrees C in a range of -20 to 50 degrees C between the thermistors when applying hardware linearization followed by software linearization using the Steinhart-Hart equation. The measurement system presented a maximum standard uncertainty of 0.17 degrees C for the entire measurement range. A maximum error of 0.3 degrees C relative to the set-point and an overshoot of around 1% were observed when the system was used in a closed-loop control using the proportional-integral strategy.

This paper presents a hybrid approach to predict the evolution of technological maturity of R&D projects, using the context of the oil and gas (O&G) sector as an example. Integrating System Dynamics (SD) and Agent-based Modeling (ABM) enables the proposed multilevel framework to capture uncertainties inherent to R&D projects, including work effort, team size, and project duration, all of which influence technological progress. Although AB-SD hybrid models are well established in other fields, their application in R&D contexts remains limited. The AB-SD model combines system-level feedback structures governing work phases, rework cycles, and project duration with the explicit representation of decentralized agents (e.g., team members, tasks, and controllers) whose interactions drive emergent project dynamics. A base-case scenario was developed to analyze the structural dynamics of early-stage innovation projects, simulating 15 parallel tasks over 156 weeks. In a comparative scenario with sequential task execution, the model showed an 88% reduction in rework duration relative to the base case. The second scenario evaluated mixed parallel-sequential task structures under varying team sizes. In parallel configuration, simulation results indicated that increasing team size reduced overall project duration and improved task completion rates, with optimal performance achieved for teams of four to five members. These outcomes are consistent with empirical observations in R&D project management, where moderate team expansion enhances coordination efficiency without incurring communication overhead. However, as widely recognized in empirical studies, a substantial increase in team size does not necessarily translate into higher completion rates, as excessive team growth often introduces communication complexity and management delays. Overall, the model outputs and the proposed modeling framework are well aligned with expert understanding in the field, confirming their validity as a quantitative tool for analyzing resource allocation, task scheduling efficiency, and technology maturity progression.

Electric submersible pump (ESP) systems are essential in the oil industry. These systems allow operation with high flow rates and efficiency, even in mature and deep wells. This paper compares the practical implementation of Model Predictive Controllers (MPC) in an ESP system in the Artificial Lift Laboratory at UFBA. The first controller is the traditional MPC, and the second is a target MPC with targets at the input. The zone controller is a more advantageous option for the scenarios tested since tuning is more straightforward, has an easy operating point for the operator to understand, and operates naturally in the maximum production region. Copyright (c) 2025 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)

This study presents a mathematical modeling framework that leverages recurrent neural networks (RNNs), specifically echo state networks (ESNs) and long short-term memory (LSTM) architectures, for the predictive modeling of electrical submersible pumps (ESPs) in offshore oil extraction operations. Utilizing real operational data from an offshore oil field, the research addresses the inherent complexity and nonlinear dynamics of ESP systems by employing these recurrent structures to capture and represent the temporal dependencies within the data. Key challenges, such as data noise, variability, and limited diversity, are systematically tackled to ensure robust dynamic modeling. A comparative analysis evaluates the performance of ESN and LSTM models under these constrained data conditions, aiming to identify the superior model in terms of predictive accuracy and resilience. The modeling approach emphasizes the formulation, parameterization, and validation processes essential for effective ESP optimization and control in real-world industrial settings. Findings reveal the distinct strengths and limitations of each RNN variant when applied to offshore operational data. Copyright (c) 2025 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)

This paper presents the first real-time implementation of a novel approach to nonlinear model predictive control (NMPC) with economic objectives, designed to maximize production in a well operated by an electrical submersible pump (ESP). The application of advanced control strategies in ESP systems has significantly expanded over the past decade, as these approaches allow wells to operate under optimal conditions. This results in increased and more stable production, ensuring compliance with process constraints and prolonging equipment lifespan. The proposed controller considers a zone control scheme to systematically accommodate the time-varying operational constraints commonly encountered in ESP systems (downthrust and upthrust). In order to improve the computational challenges of the NMPC law, an auto-regressive Echo State Network (ESNx)-oriented data-driven model is integrated into the closed-loop feedback system to predict controlled variables, utilizing historical measurements from the ESP pilot plant under study. The proposed NMPC approach is implemented in MATLAB/CasADi, and real-time communication (read and write) is achieved using an industrial communication protocol. Before its application in the real process, the tuning of the NMPC+ESNx controller parameters was empirically performed, initially within a simulation environment to allow for iterative adjustments based on the system's dynamic performance. Subsequently, this tuning was refined in the pilot plant to guarantee optimal performance in the maximum production region while ensuring computational feasibility. After 5 h of experiments, the results demonstrate economic gains, with a total accumulated production of 17 m3 of oil, while operating close to the pump's downthrust and upthrust limits. Furthermore, the average computational time of 25 ms per solution highlights the feasibility of implementing the algorithm in industrial hardware (PLC), enhancing its practicality for field operations.

This research aims to develop a simple regulatory controller to control a Core Annular Flow (CAF) in the oil and gas industry, focusing on transporting heavy oils. CAF is an economical method to transport viscous crude oil where less viscous liquid, typically water, is used to lubricate the pipe walls, creating an annular flow regime. However, managing the stability of CAF is challenging due to geometric variations, changes in pipeline flow direction, and emulsion formation. We used computational fluid dynamics (CFD) simulations to represent the CAF system and subsequently designed a simple control structure for the process. This process involved conducting both open-loop and closed-loop tests. The findings from the study indicate that the I controller significantly improves the system's response to disturbances in oil velocity by adeptly adjusting water velocity. This adjustment is crucial for maintaining the desired oil fraction and sustaining an annular flow pattern. An important observation was the effectiveness of the proportional gain in tracking the setpoint within annular flow regimes and the enhanced system stability achieved by increasing the integral action. The study concludes that the PI controller stabilizes operations in previously challenging conditions and expands the system's operational range.

This paper presents the real-time implementation of a nonlinear model predictive control (NMPC) strategy integrated with nonlinear estimation techniques, aimed at maximizing the production of an artificial lift system operated by an electrical submersible pump (ESP). The NMPC integrates a soft sensor into the control loop to estimate difficult-to-measure variables, thereby improving the accuracy of the controller’s internal model. Additionally, the strategy includes a zone control scheme to handle typical operational constraints of ESP systems, such as variations in downthrust and upthrust over time. The algorithm was developed using the open-source CasADi platform and validated in real-time on an ESP pilot plant. This work also presents an experimental comparison between the NMPC+NMHE and NMPC+EKF strategies, evaluating performance, estimation accuracy, and computational effort. The experimental results demonstrate that the proposed controller effectively maintains the pilot plant operation within a safe and feasible region while optimizing production, even under challenging conditions such as dynamic model uncertainties, parametric variations, and process noise. Both estimation strategies exhibit satisfactory performance; however, NMPC+NMHE provides lower estimation variability, whereas NMPC+EKF offers lower computational cost.

The electric submersible pump (ESP) is widely used in oil extraction processes and is recognized for its effectiveness as an artificial lift technique in the petroleum sector. Developing and improving predictive models for these systems can contribute to optimizing operational efficiency and maximizing oil production. These improvements facilitate ESP performance control and monitoring, potentially making the extraction process more economically and environmentally sustainable. This paper aims to develop a framework for creating an interpretable corrective model for the ESP phenomenological model using experimental data. Neural networks were employed to analyze the complexities and nonlinearities of the processes, followed by symbolic regression to generate a simplified and interpretable equation to improve the model's predictive capacity. An uncertainty assessment of model parameters was performed using Markov chain Monte Carlo (MCMC) and propagated to the model prediction. Additionally, a regression model was created directly from the process data for comparison purposes. The comparative analysis indicated that the approach incorporating neural networks to generate synthetic data, followed by symbolic regression, improved the model's ability to predict key variables such as intake pressure, choke pressure, and production flow.

This work proposes a new methodology to identify and validate deep learning models for artificial oil lift systems that use submersible electric pumps. The proposed methodology allows for obtaining the models and evaluating the prediction's uncertainty jointly and systematically. The methodology employs a nonlinear model to generate training and validation data and the Markov Chain Monte Carlo algorithm to assess the neural network's epistemic uncertainty. The nonlinear model was used to overcome the limitations of the need for big datasets for training deep learning models. However, the developed models are validated against experimental data after training and validation with synthetic data. The validation is also performed through the models' uncertainty assessment and experimental data. From the implementation point of view, the method was coded in Python with Tensorflow and Keras libraries used to build the neural Networks and find the hyperparameters. The results show that the proposed methodology obtained models representing both the nonlinear model's dynamic behavior and the experimental data. It provides a most probable value close to the experimental data, and the uncertainty of the generated deep learning models has the same order of magnitude as that of the nonlinear model. This uncertainty assessment shows that the built models were adequately validated. The proposed deep learning models can be applied in several applications requiring a reliable and computationally lighter model. Hence, the obtained AI dynamic models can be employed for digital twin construction, control, and optimization.

This paper presents a novel approach to digitizing and modeling pressure swing adsorption (PSA) processes using an uncertainty-aware digital twin. PSA modeling presents unique challenges due to its complex and cyclic behavior, which lacks a steady state. By contributing to the literature on periodic systems, we provide valuable insights into the potential applications of artificial intelligence and digital twins beyond the field of cyclic processes. Our proposed methodology can enhance the understanding and optimization of complex systems across various industries and applications. The proposed digital twin is uncertainty-aware and reliable, continuously updating itself through online learning and utilizing a novel feedback tracker to accurately represent the PSA system. This robust and adaptable methodology supports optimal PSA system operation and facilitates informed decision-making for enhanced process operation. The results demonstrate that the proposed approach yields a reliable digital twin for the PSA unit, capable of tracking the process’s complex dynamics and adapting to changes, including adsorbent degradation, which is a significant challenge in PSA operations. Overall, this work highlights the potential of advanced technologies, such as digital twins and artificial intelligence, to improve performance and efficiency in the field of process engineering. This work contributes to the ongoing efforts to optimize industrial processes and support sustainable development by providing a reliable and adaptable methodology for digitizing PSA processes.

This paper address the problem of docking an Autonomous Underwater Vehicle (AUV) with a mobile docking station using Nonlinear Model Predictive Control (NMPC) strategy. The optimal control goal is to track the predicted motion of the docking station over the prediction horizon, instead of using only its current states. Further, co-visibility constraints are proposed to ensure that the docking station remains in the Field Of View (FOV) of the AUV and vice-versa. These constraints would allow the AUV and the Docking Station (DS) to use a vision-based localization in the terminal phases of docking operation. The performance of the proposed predictive strategy is illustrated through simulation.

This paper proposes the implementation of model-based soft sensors to experimentally validate process variables that are difficult-to-measure in the upstream oil industry, such as fluid viscosity, productivity index, and average flow rate in the production column. The study focuses on an oil production pilot plant operated by an electrical submersible pump (ESP), which was fully instrumented with a supervisory system that collected and stored data, enabling the validation of algorithms. We evaluate the design and use of the Extended Kalman Filter (EKF), the Unscented Kalman Filter (UKF), and the Nonlinear Moving Horizon Estimator (NMHE) to monitor and estimate the process variables and compare their performance in a real environment. The results demonstrated that all three estimators EKF, UKF, and NMHE performed satisfactorily under both transient and steady-state operating conditions. However, NMHE outperformed the others by delivering lower estimation errors and better accuracy over 40 h of continuous operation at the oil well pilot plant monitored by ESP. This superior performance is quantified using the indicator root mean squared error (RMSE). With a 10-depth window, NMHE showed greater accuracy in estimating system states, achieving a global average RMSE of 0.207, in contrast to EKF’s 0.282 and UKF’s 0.280. Overall, the study highlights the potential of the evaluated model-based soft sensors in the upstream petroleum industry to measure challenging variables precisely.

The paper presents a case study that applies a model predictive control (MPC) approach in a Micro850 programmable logic controller (PLC) to a laboratory pressure swing adsorption (PSA) process used for separating gas mixtures of CO2 and CH4. PLC is an industrial hardware characterized by its robustness to hazardous environments and limited computational capacities, which poses computational challenges for MPC implementation. This paper’s main contribution is the application of the modified Takagi–Sugeno–Kang-based MPC (MTSK-MPC) algorithm to this PSA unit, which provides features to investigate and implement feasible MPC designs in PLCs. The investigation consists of a sensitivity analysis of how some design parameters influence the PLC memory and the MPC implementation and a comparative evaluation of the computational processing from different MPC algorithms and simulations. The comparison comprises software-in-the-loop simulations with three algorithms in the PC: an implicit MPC, an explicit MPC, and the MTSK-MPC. Additionally, it includes a hardware-in-the-loop simulation with the implemented MTSK-MPC in Micro850. The results show that the MPC algorithms achieve close performance, tracking setpoint changes and rejecting output disturbances, with the MTSK-MPC presenting the lower processing time among the MPCs in the PC. The study concludes that the implementation of MTSK-MPC in the Micro850 is feasible.

Robust learning is an important issue in Scientific Machine Learning (SciML). There are several works in the literature addressing this topic. However, there is an increasing demand for methods that can simultaneously consider all the different uncertainty components involved in SciML model identification. Hence, this work proposes a comprehensive methodology for uncertainty evaluation of the SciML that also considers several possible sources of uncertainties involved in the identification process. The uncertainties considered in the proposed method are the absence of a theory, causal models, sensitivity to data corruption or imperfection, and computational effort. Therefore, it is possible to provide an overall strategy for uncertainty-aware models in the SciML field. The methodology is validated through a case study developing a soft sensor for a polymerization reactor. The first step is to build the nonlinear model parameter probability distribution (PDF) by Bayesian inference. The second step is to obtain the machine learning model uncertainty by Monte Carlo simulations. In the first step, a PDF with 30,000 samples is built. In the second step, the uncertainty of the machine learning model is evaluated by sampling 10,000 values through Monte Carlo simulation. The results demonstrate that the identified soft sensors are robust to uncertainties, corroborating the consistency of the proposed approach.

Over the years, the robot grasping area has received considerable attention. Several techniques were developed to increase the efficiency and autonomy of robot manipulators. This work presents a brief review and analysis of different techniques applied to deep learning-based robot grasping generation models, in addition to some of the challenges that need to be overcome in the area.