Energy-efficient smart grid operations through dynamic digital twin models and deep learning

Research on coal mine safety management based on digital twin

In the control of complex systems, we observe two diametrical trends: model-based control derived from digital twins, and model-free control through AI. There are also attempts to bridge the gap between the two by incorporating learning-based AI algorithms into digital twins to mitigate mismatches between the digital twin model and the physical system. One of the most straightforward approaches to this is direct input adaptation. In this paper, we ask whether it is useful to employ a generic learning algorithm in such a setting, and our conclusion is “not very”. We denote an algorithm to be more useful than another algorithm based on three aspects: 1) it requires fewer data samples to reach a desired minimal performance, 2) it achieves better performance for a reasonable number of data samples, and 3) it accumulates less regret. In our evaluation, we randomly sample problems from an industrially relevant geometry assurance context and measure the aforementioned performance indicators of 16 different algorithms. Our conclusion is that blackbox optimization algorithms, designed to leverage specific properties of the problem, generally perform better than generic learning algorithms, once again finding that “there is no free lunch”. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —Digital twins have the potential to improve productivity and quality in complex systems such as manufacturing systems. Their impact on system performance hinges on the accuracy of their digital models around the system’s operating points. Difficult to measure phenomena, such as wear and tear of equipment, however, may cause a mismatch between the digital twin model and the physical system. In this paper, we formalize this problem and compare 16 potential solution strategies under practical aspects. We argue that readily available off-the-shelf blackbox optimization algorithms may prove more useful for this problem, than more recent learning-based approaches. Specifically, gradient-based algorithms will perform best in systems with high-dimensional, continuous, and non-linear performance functions – even in the presence of white measurement noise.

Modern Digital Twin models are normally built based on massive live data from the manufacturing life-cycle to realize the real-time virtual representation of the physical system and applied in predictive simulation for optimization suggestions and fault warnings for the subsequent operation. Four residual life prediction methods based on deep learning are established to build Bearing Digital Twin models, including the classical CNN and RNN, as well as the LSTM and CNN-LSTM. The real rolling bearing digital twin models are setup and predictive simulated based on the given datasets and evaluation indicators. The simulation results are analyzed and the best performance models for Bearing residual life prediction are stated as a conclusion. Some future research points on deep learning based digital twin modeling and simulation are proposed.

As an innovative information technology, Digital Twin has greatly promoted the development of intelligent manufacturing in the industry. However, in the field of geotechnical engineering, there are still few researches on this aspect, which is still a “new territory” and “no man's land”. The concept of digital twin coincides with the needs of geotechnical engineering informatization, so the introduction of digital twin technology into the field of geotechnical engineering will help to promote the development process of geotechnical engineering informatization and digitization. This paper puts forward and defines the digital twin model of geotechnical engineering, describes the connotation of the digital twin model, and studies the architecture of the digital twin model of geotechnical engineering. On this basis, the integration and sharing mechanism of geotechnical engineering digital twin data based on BIM technology is proposed. In order to break through the defect that the geotechnical engineering information model has not fully played its role for a long time, an integrated model of geological body and structural body is constructed based on the construction and integration module of geotechnical digital twin model. Furthermore, the geotechnical engineering digital twin simulation analysis module is developed to initially form the geotechnical engineering digital twin model, so as to realize the geotechnical engineering digital design, collaborative construction, visual decision-making and transparent management.

Digital twin model construction of robot and multi-object under stacking environment for grasping planning

A digital twin model of the axial temperature field of a DC cable for millisecond calculations

Digital twin aims to create a virtual model for a physical structure by combining measurement data in structural health monitoring. The most important feature is to achieve the physical structure-monitoring data synchronization. For this purpose, a physics-data hybrid framework to develop the bridge digital twin model in structural health monitoring is proposed in the paper. The physical base is firstly formed by the finite element model of the digital representation for the physical bridge that can fully incorporate both structural geometry and structural state. The data base is then built by all measurement data of the monitored bridge. By defining the context that is common to both physical base and data base, the mirror relationship between physical base and data base for the specified context is formulated. To achieve the best matching of the mirror relationship by minimizing process, the digital twin model in terms of the specified context can be developed. In such a way, the proposed framework integrates physical knowledge and data intelligence into one model. A demonstration of a simulated simply supported beam is provided to show how the digital twin model is developed by using proposed physics-data hybrid framework. It is found that the generated digital twin model is consistent with the current structural state of the beam. The presented physics-data hybrid framework helps in clearer understanding of the realization of digital twin model in structural health monitoring, providing a new perspective for smart bridge solutions.

In the process of the construction of China's new power system, we will vigorously promote the research and development of UHV power equipment and the wide application of power electronic devices. UHV power equipment has complex insulation structure and huge volume, bear impact energy load of wind power and photovoltaic of new power system for a long time. It will cost a lot to carry out on-site operation and maintenance tests. Digital twin technology is becoming more and more perfect, and new power system construction is gradually introduced from automobile, aviation and other manufacturing industries. Based on this, this paper introduces the digital twin technology into the high-end power equipment of the new power system, and carries out on-site operation and maintenance simulation test and functional response analysis under high current, high voltage and multi harmonic loads according to its twin model. From the four sensing dimensions of mechanical vibration, gas composition, optical vision and electrical parameters, the improvement of intelligent sensing technology of new power system equipment is analyzed, and the interaction between on-site operating parameters and digital twin model data is realized. On the other hand, GPU computing power expansion technology supporting digital twin multi-source sensing technology is proposed, which can effectively support the dynamic behavior simulation monitoring of equipment from 10-5 seconds to 103 seconds, and the operation life evaluation strategy of high-end equipment is proposed. This paper focuses on the 3D construction of the digital twin model of the high-end equipment of the new power system, and its research method can be extended to the construction of the whole network digital twin model of the new power system. The research results can provide theoretical guidance and technical reference for the application of digital twin technology in high-end power equipment scenarios, and effectively support the safe and stable operation of the new power system with "double high characteristics".

Operational digital twins bring together digitalization technologies, including machine learning, IIoT, data analytics, process simulation and optimization. A digital twin model reconciled with real-time process data provides the foundation for a range of layered applications that are key to transforming the way that process plants and value chains are operated and managed. After outlining the basic numerical methods, this paper will describe three industrial applications of dynamically reconciled digital twin models in gas processing, refining and olefins production. The focus is on application robustness and high availability, delivering increased operating margins and plant flexibility. An operational digital twin model can provide a versatile tool for real-time process optimization, and what-if analysis, as well as operations planning and scheduling. To be effective, the digital twin model needs to be reconciled with real-time process data frequently. In a dynamic operating environment where feed composition, process conditions and product demand may change continuously, this challenging data reconciliation task is solved using a combination of numerical methods: Equation-oriented model structures and solvers provide the flexibility required to solve simulation, parameter estimation and optimization cases. Extended Kalman filter and moving horizon estimation enable reconciliation using real-time dynamic data without the need for steady-state operation. Process models are developed using both first principles and data-driven surrogate modelling methods. A dynamically reconciled digital twin model provides the basis for an equation-oriented multi-period optimizer that can be solved for facility-wide steady state optimization, short-range dynamic optimization and long-range planning and scheduling problems within a single unified modelling and optimization framework. The use of hybrid models that include both data-driven and first principles components can significantly reduce the deployment and maintenance effort, the resulting low-order models facilitate model reuse across the different optimization horizons and objectives and enable large-scale optimization problems to be solved across an entire process plant or value chain. Three industrial applications of these methods will be outlined. The first is a gas plant optimizer that provides real-time process optimization and what-if analysis. The reconciled gas plant model tracks the process under significant feed and ambient variations. The second application is a refinery-wide optimizer that encompasses multiple crude distillation, vacuum and reaction units. The third application is an olefins process scheduling application that optimizes feed distribution, cracking furnace operating conditions and decoke cycle over an extended future time horizon.

Existing centralized building management systems suffer from high response latency and limited control capabilities. This study develops a hierarchical edge computing framework that integrates digital twin technology and model predictive control to achieve real-time, multi-objective environmental optimization. The proposed architecture consists of four distributed layers: a heterogeneous sensor network, edge preprocessing on a Raspberry Pi 4 device, digital twin modeling using Kalman filtering, and MPC optimization using the SLSQP algorithm. Multi-source data fusion integrates 44,326 records from the EPA, PhysioNet, and NOAA databases using cubic spline interpolation and PCA dimensionality reduction. The digital twin utilizes a simplified physics-based model for six-hour condition predictions, while MPC optimizes HVAC operation at 15-minute control intervals using ventilation rates and temperature setpoints. System evaluation results demonstrated an end-to-end response latency of 1.8 seconds with 71% CPU utilization on edge hardware. Prediction accuracy reached 75%, energy savings reached 10.4%, and target maintenance achieved 94.7%. Compared to PID control, this framework improved CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> management efficiency by 41.4% and reduced comfort violations by 63.2%. The distributed edge computing architecture demonstrated superior computational efficiency and economic feasibility, with a payback period of 2.2 years.

Additive manufacturing (AM) has revolutionized the manufacturing industry by offering flexibility, customization, and rapid prototyping capabilities. Complex geometries and low volume parts can be produced via AM in much lower lead time and cost compared to traditional manufacturing methods. However, ensuring the quality and reliability of AM parts remains a significant challenge due to variations in material properties and process parameters. Traditional material property analysis methods, that require experimental testing and computational modeling, are often costly and time-consuming and often not applicable to the complex geometries of AM components. These challenges are further magnified when repetitive sampling is required. To address these challenges, this paper presents a novel approach that combines experimental findings with Finite Element Modeling (FEM) to construct a digital twin (DT) model of AM parts. The DT model will serve as a virtual representation along with real-time monitoring, simulation, and optimization capabilities. Numerical simulations are conducted to correlate the resonance frequency of parts at different process conditions with their material characteristics, represented as RF Z-score values. Hence, the results suggest that such an approach provides manufacturers with cost-effective and efficient means of analyzing material properties and optimizing AM processes. Moreover, by leveraging digital twin technology, manufacturers can achieve greater process control and quality assurance in additive manufacturing. The findings of this study contribute to advancing the understanding and implementation of digital twin technology in AM; therefore, paving the way for enhanced productivity and reliability in additive manufacturing processes.

The current research proposes an energy optimization model to the production systems in reaction to the increased levels of demand concerning enhanced energy management in real-time in the industrial sector of manufacturing. The proposed framework is based on digital twin technology, which plays the role of the core enabling mechanism in the methodology. First, a simulation architecture of production lines is created, then a thorough explanation is given on how the digital twin model is designed along four different dimensions, namely geometry, physics, production behavior, and simulation rules. BP neural networks are used to create specific energy consumption models of each device to represent the energy properties of each piece of equipment. Then a method of optimization is created that targets workshop-level production processes and combines a multi-objective objective function based on several assessment criteria. They are tool life, robot motion smoothness and production time. This dynamic model is fed directly into this optimization process using energy consumption data extracted out of the digital twin model. When dealing with the collaborative adjustment of machining parameters between several machines when producing low carbon products, the artificial bee colony algorithm is applied, providing a strong global search capability, which is well suited to the complexity of this optimization problem. The proposed strategy is verified using a case study focusing on the production process of a particular workpiece. During normal operating conditions, the optimization of both workshop energy consumption and production time reduces energy use by 29.71 percent compared to the traditional approach. In rework situations, more emphasis on tool life and robot motion smoothness results in a 12.63 percent lower plant energy consumption than the current baseline. These findings indicate that the framework is effective in supporting energy monitoring and optimization in all aspects of shop floor production activities.

Digital twin modelling for optimizing the material consumption: A case study on sustainability improvement of thermoforming process

This paper presents an autonomous drilling advisory system powered by digital twins and AI solutions. Such an advisory system aims to automate real-time monitoring and parameter optimization, reduce subject-matter experts, and meet the demands for safer and more efficient drilling toward autonomous operation. The methodology proposed in this research involves the creation of a comprehensive Digital Twin model that accurately replicates the drilling process by integrating hydraulic, thermal dynamic, and mechanical models. To ensure high model accuracy, an auto-calibration approach is developed, driven by real-time data, to fine-tune the Digital Twin models. Additionally, AI-based model reasoning techniques are employed to detect potential hazards and risks ahead of the bit proactively. This is achieved by comparing the ideal behavior of the digital twin replica with the actual behavior observed from downhole and the rig. As a result, real-time diagnostics are generated to supervise ongoing operations, accompanied by suggestions to mitigate identified risks. Furthermore, the system leverages the capabilities of the Digital Twin and optimization methods to create multiple combinations of operational parameters. These parameters are optimized by ranking the predicted performance derived from the Digital Twin. The optimized operational parameters are automatically generated as forward advice to drillers, enabling them to make informed decisions and enhance drilling performance. Testing results on multiple wells from different operators are presented, showcasing the system's capabilities in real-time monitoring and drilling parameter optimization. The system demonstrates its effectiveness in providing diagnostic messages with early anomaly detection during drilling and casing running. These diagnostic warnings include losses, leakage, poor hole cleaning, and stuck pipe, enabling proactive intervention to mitigate risks. Furthermore, the system optimizes operational parameters during drilling and tripping in real-time without requiring human intervention. This optimization covers parameters such as flow rate, rotary speed (RPM), and rate of penetration (ROP) during drilling, and tripping speed during tripping in and pulling out of the hole. The time savings achieved through the use of optimized parameters are quantified for both cases, demonstrating a substantial improvement in operational efficiency while maintaining safety margins. The scalability and adaptability of the system are also highlighted, emphasizing its ability to accommodate diverse drilling scenarios and integrate with existing solutions in various deployment conditions. The proposed methodology demonstrates the development of a robust and efficient system that enhances decision-making and improves drilling performance. In addition, the results highlight the potential benefits of combining AI and Digital Twin technologies in the drilling industry, paving the way for future innovations and advancements in the field.

We develop novel approaches to enhance the functionality and efficiency of digital twins through deep learning techniques. Digital twins, sophisticated virtual models of physical systems, serve as dynamic counterparts, mirroring real-world entities' behavior and performance. Their primary role includes real-time monitoring, flaw detection, automated control, and proactive maintenance. To perform such tasks it is necessary to use various methods from scientific computing to combine data with physics models. Furthermore, to assess the reliability of a digital twin, it is important to not only make predictions, but also quantify the uncertainty of estimations and predictions. In this regard, Bayesian methods serve as a natural framework to pose the problems. However, the complexity and real-time operation requirements pose significant computational challenges. The thesis explores the integration of deep learning in scientific computing for enabling digital twins. The focus is on overcoming the computational hurdles in real-time data assimilation and inverse problem-solving through surrogate modeling. Deep learning, with its capability to handle high-dimensional functions in both deterministic and stochastic settings is shown to be a viable solution to these challenges. This work investigates several key areas, beginning with a brief introduction to digital twins, data assimilation, and inverse problems. It proceeds to elaborate on the role of Bayesian inversion problems in static and dynamic contexts, highlighting the importance of uncertainty quantification. The focus is on the challenges of real-time computation and the potential of Bayesian methods, despite their computational intensity. The thesis investigates deep learning architectures like dense and residual neural networks, convolutional neural networks, recurrent neural networks and transformers. The core contributions are presented in the four main chapters of the thesis: 1) Reduced Order Modeling for Parameterized Time-Dependent Partial Differential Equations: Introducing a deep learning framework for solving parameterized PDEs using spatially and memory-aware neural networks, demonstrated on linear advection and incompressible Navier-Stokes equations. 2) Markov Chain Generative Adversarial Neural Networks: Enhancing Bayesian inversion for static problems with sparse observations using a unique algorithm combining MCMC and GANs, demonstrated on a Darcy flow problem and pipe flow leakage detection. 3) Probabilistic Digital Twin for Leak Localization: Developing a probabilistic framework using supervised Wasserstein Autoencoders for leak localization in water distribution networks, tested on several network models. 4) The Deep Latent Space Particle Filter: Proposing a real-time nonlinear data assimilation method for dynamic PDEs with uncertainty quantification, using transformer-based dimensionality reduction and time-stepping in particle filters, demonstrated on various dynamic systems. This thesis contributes significantly to digital twins, addressing computational limitations and opening new avenues for practical applications in various industries.