Abstract Robotic manipulation inherently involves contact with objects for task accomplishment. Traditional motion planning techniques, while having shown their success in collision-free scenarios, may not handle manipulation tasks effectively because they typically avoid contact. Although geometric constraints have been introduced into classical motion planners for tasks that involve interactions, they still lack the capability to fully incorporate contact. In addition, these planning methods generally do not operate on objects that cannot be directly controlled. In this work, building on a recently proposed framework for energy-based quasi-static manipulation, we propose an approach to manipulation planning by adapting a numerical continuation algorithm to compute the equilibrium manifold (EM), which is implicitly derived from physical laws. By defining a manipulation potential energy function that captures interaction and natural potentials, the numerical continuation approach is integrated with adaptive ordinary differential equations that converge to the EM. This allows discretizing the implicit manifold as a graph with a finite set of equilibria as nodes interconnected by weighted edges defined via a haptic metric. The proposed framework is evaluated with an inverted pendulum task, where the explored branch of the manifold demonstrates effectiveness.

Abstract Heating, Ventilation, and Air Conditioning (HVAC) systems are major energy consumers in buildings, challenging the balance between efficiency and occupant comfort. While prior research explored generative AI for HVAC control in simulations, real-world validation remained scarce. This study addresses this gap by designing, deploying, and evaluating “Office-in-the-Loop,” a novel cyber-physical system leveraging generative AI within an operational office setting. Capitalizing on multimodal foundation models and Agentic AI, our system integrates real-time environmental sensor data (temperature, occupancy, etc.), occupants’ subjective thermal comfort feedback, and historical context as input prompts for the generative AI to dynamically predict optimal HVAC temperature setpoints. Extensive real-world experiments demonstrate significant energy savings (up to 47.92%) while simultaneously improving comfort (up to 26.36%) compared to baseline operation. Regression analysis confirmed the robustness of our approach against confounding variables like outdoor conditions and occupancy levels. Furthermore, we introduce Data-Driven Reasoning using Agentic AI, finding that prompting the AI for data-grounded rationales significantly enhances prediction stability and enables the inference of system dynamics and cost functions, bypassing the need for traditional reinforcement learning paradigms. This work bridges simulation and reality, showcasing generative AI’s potential for efficient, comfortable building environments and indicating future scalability to large systems like data centers.

Abstract The principal function of an open recirculating system (ORS) is to remove heat from power plant equipment. In particular, the presence of scale on the internal surfaces of ORS heat exchange equipment can reduce heat transfer efficiency, which leads to increased energy consumption and operating costs. The purpose of this article is to investigate the process of calcium carbonate (CaCO3) precipitation formation in terms of the components of the carbonate system and parameters affecting the shift of carbonate equilibrium in an ORS. An appraisal model was used to represent the processes occurring during the operation of an ORS. In this study, it is demonstrated that water heating in ORS condensers leads to the excretion of carbon dioxide (CO2) from the water, while cooling in the cooling towers results in CO2 uptake by the water. These processes significantly influence the state of carbonate equilibrium within the ORS. The study used the results of chemical control of the make-up and cooling water at the ORS Rivne Nuclear Power Plant (RNPP) for 2022. Furthermore, the dependencies of changes in the components of the carbonate system on the pH levels of the make-up (pH 7.51–9.52) and cooling (pH 8.21–9.53) water were revealed, and changes in the cycles of concentration (CоC), total hardness (TH), total dissolved solids (TSD), and total alkalinity (TA) were estimated. Taking into account the obtained correlation dependencies, in general, it was found that the lower the CoC levels, the lower the TA reduction value, and it is possible to increase or decrease the cooling water pH levels, which is determined by the initial state of carbonate equilibrium of make-up water. These findings enable the prediction and control of CaCO3 scale formation through continuous monitoring of water chemistry, making the process more efficient, reliable, and sustainable. The results emphasize the importance of data-driven modeling for optimizing water treatment and reducing operational costs in power plants by reducing CaCO3 scale formation.

Abstract Thermal integrity profiling (TIP) is a nondestructive testing technique that takes advantage of the concrete heat of hydration (HoH) to detect inclusions during the casting process. This method is becoming more popular due to its ease of application, as it can be used to predict defects in most concrete foundation structures requiring only the monitoring of temperatures. Despite its advantages, challenges remain with regard to data interpretation and analysis, as temperature is only known at discrete points within a given cross-section. This study introduces a novel method for the interpretation of TIP readings using neural networks. Training data are obtained through numerical finite element simulation spanning an extensive range of soil, concrete, and geometrical parameters. The developed algorithm first classifies concrete piles, establishing the presence or absence of defects. This is followed by a regression algorithm that predicts the defect size and its location within the cross-section. In addition, the regression model provides reliable estimates for the reinforcement cage misalignment and concrete hydration parameters. To make these predictions, the proposed methodology only requires temperature data in the form standard in TIP, so it can be seamlessly incorporated within the TIP workflows. This work demonstrates the applicability and robustness of machine learning algorithms in enhancing nondestructive TIP testing of concrete foundations, thereby improving the safety and efficiency of civil engineering projects.

Abstract Numerical solutions of partial differential equations require expensive simulations, limiting their application in design optimization, model-based control, and large-scale inverse problems. Surrogate modeling techniques aim to decrease computational expense while retaining dominant solution features and characteristics. Existing frameworks based on convolutional neural networks and snapshot-matrix decomposition often rely on lossy pixelization and data-preprocessing, limiting their effectiveness in realistic engineering scenarios. Recently, coordinate-based multilayer perceptron networks have been found to be effective at representing 3D objects and scenes by regressing volumetric implicit fields. These concepts are leveraged and adapted in the context of physical-field surrogate modeling. Two methods toward generalization are proposed and compared: design-variable multilayer perceptron (DV-MLP) and design-variable hypernetworks (DVH). Each method utilizes a main network which consumes pointwise spatial information to provide a continuous representation of the solution field, allowing discretization independence and a decoupling of solution and model size. DV-MLP achieves generalization through the use of a design-variable embedding vector, while DVH conditions the main network weights on the design variables using a hypernetwork. The methods are applied to predict steady-state solutions around complex, parametrically defined geometries on non-parametrically-defined meshes, with model predictions obtained in less than a second. The incorporation of random Fourier features greatly enhanced prediction and generalization accuracy for both approaches. DVH models have more trainable weights than a similar DV-MLP model, but an efficient batch-by-case training method allows DVH to be trained in a similar amount of time as DV-MLP. A vehicle aerodynamics test problem is chosen to assess the method’s feasibility. Both methods exhibit promising potential as viable options for surrogate modeling, being able to process snapshots of data that correspond to different mesh topologies.

Abstract In recent years, the manufacturing sector has seen an influx of artificial intelligence applications, seeking to harness its capabilities to improve productivity. However, manufacturing organizations have limited understanding of risks that are posed by the usage of artificial intelligence, especially those related to trust, responsibility, and ethics. While significant effort has been put into developing various general frameworks and definitions to capture these risks, manufacturing and supply chain practitioners face difficulties in implementing these and understanding their impact. These issues can have a significant effect on manufacturing companies, not only at an organization level but also on their employees, clients, and suppliers. This paper aims to increase understanding of trustworthy, responsible, and ethical Artificial Intelligence challenges as they apply to manufacturing and supply chains. We first conduct a systematic mapping study on concepts relevant to trust, responsibility and ethics and their interrelationships. We then use a broadened view of a machine learning lifecycle as a basis to understand how risks and challenges related to these concepts emanate from each phase in the lifecycle. We follow a case study driven approach, providing several illustrative examples that focus on how these challenges manifest themselves in actual manufacturing practice. Finally, we propose a series of research questions as a roadmap for future research in trustworthy, responsible and ethical artificial intelligence applications in manufacturing, to ensure that the envisioned economic and societal benefits are delivered safely and responsibly.

Abstract Digital twins are a new paradigm for our time, offering the possibility of interconnected virtual representations of the real world. The concept is very versatile and has been adopted by multiple communities of practice, policymakers, researchers, and innovators. A significant part of the digital twin paradigm is about interconnecting digital objects, many of which have previously not been combined. As a result, members of the newly forming digital twin community are often talking at cross-purposes, based on different starting points, assumptions, and cultural practices. These differences are due to the philosophical world-view adopted within specific communities. In this paper, we explore the philosophical context which underpins the digital twin concept. We offer the building blocks for a philosophical framework for digital twins, consisting of 21 principles that are intended to help facilitate their further development. Specifically, we argue that the philosophy of digital twins is fundamentally holistic and emergentist. We further argue that in order to enable emergent behaviors, digital twins should be designed to reconstruct the behavior of a physical twin by “dynamically assembling” multiple digital “components”. We also argue that digital twins naturally include aspects relating to the philosophy of artificial intelligence, including learning and exploitation of knowledge. We discuss the following four questions (i) What is the distinction between a model and a digital twin? (ii) What previously unseen results can we expect from a digital twin? (iii) How can emergent behaviours be predicted? (iv) How can we assess the existence and uniqueness of digital twin outputs?

Abstract Pharmaceutical distribution routing problem is a key problem for pharmaceutical enterprises, since efficient schedules can enhance resource utilization and reduce operating costs. Meanwhile, it is a complicated combinatorial optimization problem. Existing research mainly focused on delivery route lengths or distribution costs minimization, while seldom considered customer priority and carbon emissions simultaneously. However, considering the customer priority and carbon emissions simultaneously will not only help to enhance customer satisfaction, but also help to reduce the carbon emissions. In this article, we consider the customer priority and carbon emission minimization simultaneously in the pharmaceutical distribution routing problem, the corresponding problem is named pharmaceutical distribution routing problem considering customer priority and carbon emissions. A corresponding mathematical model is formulated, the objectives of which are minimizing fixed cost, refrigeration cost, fuel consumption cost, carbon emission cost, and penalty cost for violating time windows. Moreover, a hybrid genetic algorithm (HGA) is proposed to solve the problem. The framework of the proposed HGA is genetic algorithm (GA), where an effective local search based on variable neighborhood search (VNS) is specially designed and incorporated to improve the intensification abilities. In the proposed HGA, crossover with adaptive probability and mutation with adaptive probability are utilized to enhance the algorithm performance. Finally, the proposed HGA is compared with four optimization algorithms, and experimental results have demonstrated the effectiveness of the HGA.

Abstract The “digital twin” is now a recognized core component of the Industry 4.0 journey, helping organizations to understand their complex processes, resources and data to provide insight, and help optimize their operations. Despite this, there are still multiple definitions and understandings of what a digital twin is; all of which has led to a “mysticism” around the concept. Following the “hype curve” model, it can be seen that digital twins have moved past their initial hype phase with only minimal implementation in industry, this is often due to the perceived high cost of initial development and sensor outfit. However, a second hype peak is predicted through the development of “lean digital twins.” Lean digital twins represent conceptual or physical systems in much lower detail (and hence at much lower cost to build and manage the models), focusing in on the key parameters and operators that most affect the desired optimal outcomes of the physical system. These lean digital twins are requirements managed with the system to ensure added value and tapping into existing architectures such as onboard platform management systems to minimize costs. This article was developed in partnership between BMT and Siemens to demystify the different definitions and components of a lean digital twin and discuss the process of implementing a lean digital twin solution that is tied to the core benefits in question and outlining the tools available to make implementation a reality.

Abstract We develop and demonstrate a computationally cheap framework to identify optimal experiments for Bayesian inference of physics-based models. We develop the metrics (i) to identify optimal experiments to infer the unknown parameters of a physics-based model, (ii) to identify optimal sensor placements for parameter inference, and (iii) to identify optimal experiments to perform Bayesian model selection. We demonstrate the framework on thermoacoustic instability, which is an industrially relevant problem in aerospace propulsion, where experiments can be prohibitively expensive. By using an existing densely sampled dataset, we identify the most informative experiments and use them to train the physics-based model. The remaining data are used for validation. We show that, although approximate, the proposed framework can significantly reduce the number of experiments required to perform the three inference tasks we have studied. For example, we show that for task (i), we can achieve an acceptable model fit using just 2.5% of the data that were originally collected.