Physics-based Approaches Research Articles

A novel hypergraph convolution network-based approach for predicting the material removal rate in chemical mechanical planarization

The material removal rate (MRR) plays a critical role in the chemical mechanical planarization (CMP) process in the semiconductor industry. Many physics-based and data-driven approaches have been proposed to-date to predict the MRR. Nevertheless, most of them neglect the underlying equipment structure containing essential interaction mechanisms among different components. To fill the gap, this paper proposes a novel hypergraph convolution network (HGCN) based approach for predicting MRR in the CMP process. The main contributions include: (1) a generic hypergraph model to represent the interrelationships of complex equipment; and (2) a temporal-based prediction approach to learn the complex data correlation and high-order representation based on the hypergraph. To validate the effectiveness of the proposed approach, a case study is conducted by comparing with other cutting-edge models, of which it outperforms in several metrics. It is envisioned that this research can also bring insightful knowledge to similar scenarios in the manufacturing process.

Satellite-derived bathymetry for maritime archaeology: Testing its effectiveness at two ancient harbours in the Eastern Mediterranean

High-resolution bathymetry is an essential tool in maritime archaeology, often used for underwater prospection, imaging, palaeo-landscape reconstruction and change detection. The commonest acquisition method relies on shipborne acoustic systems (e.g. multibeam echosounder). However, most of the world’s seabed has not been surveyed using these methods. Derivation of bathymetry from high-resolution multispectral satellite imagery (Satellite-Derived Bathymetry: SDB) can redress this in shallow (<20–30 m depth) clear water. It can be derived empirically using depth control points, or independent of control data through physics-based approaches. Given global satellite coverage, it is applicable almost anywhere with suitable water conditions and is becoming increasingly widespread in the hydrographic survey community. However, archaeological application of SDB is almost totally un-investigated. This paper examines the utility of SDB for maritime archaeology through 2 case studies: the ancient harbours of Tyre (Lebanon) and Ras Ibn Hani (Syria). Tested data comprise 2 m-resolution physics-based commercial SDB and medium-resolution (10–30 m) SDB empirically derived from open access multispectral imagery (Landsat 8, Sentinel-2). SDB was assessed for its ability to detect submerged cultural features and landforms and as input into palaeo-geographic reconstructions. Tested SDB performed poorly for feature detection, but was better-suited to landform identification and palaeo-geographic reconstruction. SDB also has clear advantages: low cost, wide area coverage, speed and data availability. Disadvantages include variable accuracy, technical limitations (e.g. need for clear water) and approach-specific requirements (e.g. depth control points for empirical and complex modelling for physics-based approaches). For maritime archaeology, its current limitations means that it will not replace acoustics for prospection and imaging but it does have utility for submerged landscape investigation and reconstruction, particularly in areas which now only have no/poor bathymetric data.

Metaheuristic least squares support vector machine-based lateral strength modelling of reinforced concrete columns subjected to earthquake loads

Existing physics-based methods for estimating the lateral strength of reinforced concrete (RC) columns lack generalization capabilities, where the methods for predicting the shear strength of shear- and flexure-shear-critical columns cannot be used to predict the flexural strength of flexure-critical columns, and vice versa. This paper proposes a nature-inspired, metaheuristic, least squares-support vector machine for regression (LS-SVMR) model for enhancing the generalization of lateral strength prediction of RC columns. The metaheuristic LS-SVMR model is a hybrid computational intelligence approach, which integrates LS-SVMR to elucidate the nonlinear mapping between influential factors and the lateral strength with a nature-inspired metaheuristic optimization algorithm, called coupled simulated annealing (CSA), to efficiently tune the hyper-parameters and facilitate the training process. The generalization performance is validated by comparison with several machine learning (ML) models using cross-validation strategies and with commonly used physics-based approaches as well as by successfully modeling the physical processes associated with variation of lateral strength of flexure-, shear-, and flexure-shear-critical columns when input predictors vary.

Dual-Wavelength Polarimetric Lidar Observations of the Volcanic Ash Cloud Produced during the 2016 Etna Eruption

Lidar observations are very useful to analyse dispersed volcanic clouds in the troposphere mainly because of their high range resolution, providing morphological as well as microphysical (size and mass) properties. In this work, we analyse the volcanic cloud of 18 May 2016 at Mt. Etna, in Italy, retrieved by polarimetric dual-wavelength Lidar measurements. We use the AMPLE (Aerosol Multi-Wavelength Polarization Lidar Experiment) system, located in Catania, about 25 km from the Etna summit craters, pointing at a thin volcanic cloud layer, clearly visible and dispersed from the summit craters at the altitude between 2 and 4 km and 6 and 7 km above the sea level. Both the backscattering and linear depolarization profiles at 355 nm (UV, ultraviolet) and 532 nm (VIS, visible) wavelengths, respectively, were obtained using different angles at 20°, 30°, 40° and 90°. The proposed approach inverts the Lidar measurements with a physically based inversion methodology named Volcanic Ash Lidar Retrieval (VALR), based on Maximum-Likelihood (ML). VALRML can provide estimates of volcanic ash mean size and mass concentration at a resolution of few tens of meters. We also compared those results with two methods: Single-variate Regression (SR) and Multi-variate Regression (MR). SR uses the backscattering coefficient or backscattering and depolarization coefficients of one wavelength (UV or VIS in our cases). The MR method uses the backscattering coefficient of both wavelengths (UV and VIS). In absence of in situ airborne validation data, the discrepancy among the different retrieval techniques is estimated with respect to the VALR ML algorithm. The VALR ML analysis provides ash concentrations between about 0.1 μg/m3 and 1 mg/m3 and particle mean sizes of 0.1 μm and 6 μm, respectively. Results show that, for the SR method differences are less than &lt;10%, using the backscattering coefficient only and backscattering and depolarization coefficients. Moreover, we find differences of 20–30% respect to VALR ML, considering well-known parametric retrieval methods. VALR algorithms show how a physics-based inversion approaches can effectively exploit the spectral-polarimetric Lidar AMPLE capability.

Computational Models and Simulations of Cancer Metastasis

The dawn of the new era in precise diagnostic devices and personalized cancer treatment is intertwined with computational models capable of integrating biochemical factors and biophysical processes to simulate and predict cancer progression. In the last decade, thanks to the increase in the computational power, the development of more sophisticated and realistic models has gained attention in cancer research. These computational models can advance our fundamental understandings of the complicated processes involved in metastasis, as a challenging stage for cancer treatment, and can provide new means for developing novel tools for predicting cancer progression. Considering the potential of these models and the plethora of recent computational models of different steps of the metastasis process, this timely review provides an up-to-date outlook of novel approaches, identifies research gaps and suggests future research directions. This review focuses on physics-based approaches for modeling metastasis process and covers recent computational models and frameworks related to all steps of metastasis from primary tumor growth to secondary tumor formation. In this review, computational models and simulations are classified based on their targeted step of metastasis. Tumor growth and tumor-induced angiogenesis together are considered as a necessary primary step for the metastasis cascade that has four steps: the intravasation of an individual tumor cell into circulatory system, circulation of the cell, arrest and extravasation, and eventually colonization and formation of a metastatic tumor.

Temporal logic learning-based anomaly detection in metroplex terminal airspace operations

The airspace system is a complex dynamical system with complicated controlled interactions between its constituent subsystems – terminal airspace, en-route airspace, and ground. Of these, air traffic management in the multi-airport (metroplex) terminal airspace is one of the most complicated subsystems to manage, especially due to the interactions between proximal airports. Analyzing anomalous behaviors in the metroplex is emerging as a key problem in understanding air traffic management complexity and safety. Although physics-based approaches have been studied in-depth for this application, newfound interest has been observed to use recorded time-series air traffic surveillance and airport operations datasets for this purpose. In this paper, we propose a machine learning-based anomaly detection algorithm that generates mathematical models to detect anomalies in metroplex operations. Several machine learning algorithms have been developed to detect anomalies using only air traffic surveillance data, but there is a significant scope of improvement by including airport operational characteristics as well, since integrating such closely-controlled metroplex operational datasets allows the developed models to effectively detect anomalies. The key contribution of this paper is in allowing anomaly detection models to recursively update so as to adapt to changes in metroplex operations. The proposed algorithm is demonstrated with real air traffic surveillance and airport operations datasets at LaGuardia, John F. Kennedy, and Newark airports, thereby detecting anomalies for operations in the New York metroplex.

Combustion Tuning for a Gas Turbine Power Plant Using Data-Driven and Machine Learning Approach

Abstract Conventional physics-based or experimental-based approaches for gas turbine combustion tuning are time consuming and cost intensive. Recent advances in data analytics provide an alternative method. In this paper, we present a cross-disciplinary study on the combustion tuning of an F-class gas turbine that combines machine learning with physics understanding. An artificial-neural-network-based (ANN) model is developed to predict the combustion performance (outputs), including NOx emissions, combustion dynamics, combustor vibrational acceleration, and turbine exhaust temperature. The inputs of the ANN model are identified by analyzing the key operating variables that impact the combustion performance, such as the pilot and the premixed fuel flow, and the inlet guide vane angle. The ANN model is trained by field data from an F-class gas turbine power plant. The trained model is able to describe the combustion performance at an acceptable accuracy in a wide range of operating conditions. In combination with the genetic algorithm, the model is applied to optimize the combustion performance of the gas turbine. Results demonstrate that the data-driven method offers a promising alternative for combustion tuning at a low cost and fast turn-around.

A comparative study of data-driven and physics-based gas turbine fault recognition approaches

The present paper compares the fault recognition capabilities of two gas turbine diagnostic approaches: data-driven and physics-based (a.k.a. gas path analysis, GPA). The comparison takes into consideration two differences between the approaches, the type of diagnostic space and diagnostic decision rule. To that end, two stages are proposed. In the first one, a data-driven approach with an artificial neural network (ANN) that recognizes faults in the space of measurement deviations is compared with a hybrid GPA approach that employs the same type of ANN to recognize faults in the space of estimated fault parameter. Different case studies for both anomaly detection and fault identification are proposed to evaluate the diagnostic spaces. They are formed by varying the classification, type of diagnostic analysis, and deviation noise scheme. In the second stage, the original GPA is reconstructed replacing the ANN with a tolerance-based rule to make diagnostic decisions. Here, two aspects are under analysis: the comparison of GPA classification rules and whole approaches. The results reveal that for simple classifications both spaces are equally accurate for anomaly detection and fault identification. However, for complex scenarios, the data-driven approach provides on average slightly better results for fault identification. The use of a hybrid GPA with ANN for a full classification instead of an original GPA with tolerance-based rule causes an increase of 12.49% in recognition accuracy for fault identification and up to 54.39% for anomaly detection. As for the whole approach comparison, the application of a data-driven approach instead of the original GPA can lead to an improvement of 12.14% and 53.26% in recognition accuracy for fault identification and anomaly detection, respectively.

A Novel Building Temperature Simulation Approach Driven by Expanding Semantic Segmentation Training Datasets with Synthetic Aerial Thermal Images

Multi-sensor imagery data has been used by researchers for the image semantic segmentation of buildings and outdoor scenes. Due to multi-sensor data hunger, researchers have implemented many simulation approaches to create synthetic datasets, and they have also synthesized thermal images because such thermal information can potentially improve segmentation accuracy. However, current approaches are mostly based on the laws of physics and are limited to geometric models’ level of detail (LOD), which describes the overall planning or modeling state. Another issue in current physics-based approaches is that thermal images cannot be aligned to RGB images because the configurations of a virtual camera used for rendering thermal images are difficult to synchronize with the configurations of a real camera used for capturing RGB images, which is important for segmentation. In this study, we propose an image translation approach to directly convert RGB images to simulated thermal images for expanding segmentation datasets. We aim to investigate the benefits of using an image translation approach for generating synthetic aerial thermal images and compare those approaches with physics-based approaches. Our datasets for generating thermal images are from a city center and a university campus in Karlsruhe, Germany. We found that using the generating model established by the city center to generate thermal images for campus datasets performed better than using the latter to generate thermal images for the former. We also found that using a generating model established by one building style to generate thermal images for datasets with the same building styles performed well. Therefore, we suggest using training datasets with richer and more diverse building architectural information, more complex envelope structures, and similar building styles to testing datasets for an image translation approach.

A Study Toward Appropriate Architecture of System-Level Prognostics: Physics-Based and Data-Driven Approaches

A Study Toward Appropriate Architecture of System-Level Prognostics: Physics-Based and Data-Driven Approaches

Prediction assessment methodology for maintenance applications in manufacturing

The aim of this paper is to present a methodology for the assessment of prognosis quality and more specifically, the Remaining Useful Life prediction in maintenance applications. The utilization of model-driven physics-based approaches in combination with prognostics is a common practice to predict a machine’s future health status. Due to the increased utilization of this concept, a methodology to estimate the prediction error is required. Scope of the paper is to present a prediction assessment methodology. The proposed concept is also validated through a predictive maintenance application in robotics.

ROM-Based Surrogate Systems Modeling of EBR-II

The System Analysis Module (SAM), developed and maintained by Argonne National Laboratory, is designed to provide whole-plant transient safety analysis capabilities for a number of advanced non–light water reactors, including sodium-cooled fast reactor (SFR), lead-cooled fast reactor (LFR), and molten salt reactor (MSR)/fluoride-salt-cooled high-temperature reactor (FHR) designs. SAM is primarily constructed as a systems-level analysis tool, with the potential to incorporate reduced order models from three-dimensional computational fluid dynamics (CFD) simulations to improve characterization of complex, multidimensional physics. It is recognized that the computational expense associated with CFD can be intractable for various engineering analyses, such as uncertainty quantification, inference, and design optimization. This paper explores the reducibility of a SAM model using recent advances in randomized linear algebra techniques, which attempt to find recurring patterns in the various realizations generated by a model after randomly perturbing all its input parameters. The reduction is described in terms of fewer degrees of freedom (DOFs), referred to as the active DOFs, for the model variables such as input model parameters and model responses. The results indicate that there is significant room for additional reduction that may be leveraged for additional computational gains when employing SAM for engineering-intensive analyses that require repeated model executions. Different from physics-based reduction approaches, the proposed approach allows one to estimate upper bounds on the reduction errors, which are rigorously developed in this work. Finally, different methods for surrogate model construction, such as regression and neural network–based training, are employed to correlate the input and output active DOFs, which are related back to the original variables using matrix-based linear transformations.

Ocean wave energy forecasting using optimised deep learning neural networks

Ocean renewable energy is a promising inexhaustible source of renewable energy, with an estimated harnessing potential of approximately 337 GW worldwide, which could re-shape the power generation mix. As with other sources of renewables, however, wave energy has an intermittent and irregular nature, which is a major concern for power system stability. Consequently, in order to integrate wave energy into power grids, it must be forecasted. This paper proposes using optimised deep learning neural networks to forecast the wave energy flux, and other wave parameters. In particular, we use moth-flame optimisation as the central decision-making unit to configure the deep neural network structure and the proper input data selection. Besides, the moth-flame optimisation algorithm was modified to improve its search space mechanisms. The forecasting skills are assessed using 13 datasets from locations across the Pacific and Atlantic coasts, and the Gulf of Mexico. The proposed optimised deep neural network performs well at all the sites, especially over short-term horizons, where it outperforms statistical and physics-based approaches.

Physics-informed machine learning for composition – process – property design: Shape memory alloy demonstration

Machine learning (ML) is shown to predict new alloys and their performances in a high dimensional, multiple-target-property design space that considers chemistry, multi-step processing routes, and characterization methodology variations. A physics-informed featured engineering approach is shown to enable otherwise poorly performing ML models to perform well with the same data. Specifically, previously engineered elemental features based on alloy chemistries are combined with newly engineered heat treatment process features. The new features result from first transforming the heat treatment parameter data as it was previously recorded using nonlinear mathematical relationships known to describe the thermodynamics and kinetics of phase transformations in alloys. The ability of the ML model to be used for predictive design is validated using blind predictions. Composition - process - property relationships for thermal hysteresis of shape memory alloys (SMAs) with complex microstructures created via multiple melting-homogenization-solutionization-precipitation processing stage variations are captured, in addition to the mean transformation temperatures of the SMAs. The quantitative models of hysteresis exhibited by such highly processed alloys demonstrate the ability for ML models to design for physical complexities that have challenged physics-based modeling approaches for decades.

Toward a Big Data-Based Approach: A Review on Degradation Models for Prognosis of Critical Infrastructure

Abstract Safety and reliability of large critical infrastructure such as long-span bridges, high-rise buildings, nuclear power plants, high-voltage transmission towers, rotating machinery, and so on, are important for a modern society. Research on reliability and safety analysis started with a “small data” problem dealing with relative scarce lifetime or failure data. Later, degradation modeling that uses performance deterioration, or, condition data collected from in-service inspections or online health monitoring became an important tool for reliability prediction and maintenance planning of highly reliable engineering systems. Over the past decades, a large number of degradation models have been developed to characterize and quantify the underlying degradation mechanism using direct and indirect measurements. Recent advancements in artificial intelligence, remote sensing, big data analytics, and Internet of things are making far-reaching impacts on almost every aspect of our lives. The effect of these changes on the degradation modeling, prognosis, and safety management is interesting questions to explore. This paper presents a comprehensive, forward-looking review of the various degradation models and their practical applications to damage prognosis and management of critical infrastructure. The degradation models are classified into four categories: physics-based, knowledge-based, data-driven, and hybrid approaches.

Integrating data-driven and physics-based approaches to characterize failures of interdependent infrastructures

Interdependent critical infrastructures (ICIs) that could trigger cascading impacts on one another have attracted great attention from academia. Most of the current studies regarding ICIs solely apply a data-driven method (e.g., inoperability input-output method) or a physics-based approach (e.g., topology network) to uniformly characterize disparate infrastructures. This manipulation not only often encounter the problem of lacking empirical data or physical knowledge as required by a designated method for depicting certain infrastructures, but also it may significantly ignore the heterogeneities of infrastructures in terms of operation mechanisms, failure characteristics, and measurement indicators. These challenges debilitate the adoption of a consistent model to characterize different infrastructures, hence a framework that can agilely integrate diverse data-driven methods and physics-based approaches is proposed to help demystify the failure propagation among interdependent infrastructures. The proposed framework consists of three major modules, including (i) understanding the basic profiles of each target infrastructure, (ii) selecting an applicable data-driven method or physics-based model to characterize each infrastructure system, and (iii) designing interfaces to operationalize the interdependencies between different infrastructures and to connect the selected methods through associated variables or parameters. Taking the geographically interdependent water supply system and road transport network as a case, the practicability of the proposed framework is demonstrated. The failure patterns of the water supply system in Hong Kong, including the hotspots, high-incidence time, and potential consequence types, were derived by mining news related to water pipe burst incidents. The cascading impacts caused by water pipe bursts on the road transport network, such as performance degradations revealed by traffic densities and vehicle delays, were further captured through physical traffic flow simulation. The developed integration framework shall contribute to the existing ICI research domain by increasing the flexibility of selecting applicable methods for individual infrastructures depending upon the availability of real-world data and physical models. Moreover, it could facilitate the retention and exploration of infrastructures’ heterogeneous features as desired by decision-makers in specific ICI research scenarios.

Physics Based Degradation Models for Electrolytic Capacitor Prognostics under Thermal Overstress Conditions

Electronics subsystems play an increasingly important role in safety critical systems for monitoring, control, and enhanced functionality. Electrolytic capacitors are an important component in ,amy subsystems that range from power supplies on safety critical avionics equipment to power drivers for electro-mechanical actuators. These capacitors are ideal forpassing or bypassing low-frequency signals in power supplies but are known to have lower reliability compared to ceramic and tantalum capacitors, and given their criticality in electronicssubsystems they are a good candidate for component level monitoring and prognostics. Prognostics provides a way to assess remaining useful life of components and systems based on their current state of health and their anticipated future use and operating conditions. Past experiences have shown that capacitor degradation and failures are quite prevalent under high electrical and thermal stress conditions that they experience during operations. Our focus in this work is on deriving a physics-based degradation model for electrolytic capacitors under thermal stress conditions. As part of our methodology, we study the effects of accelerated aging due to thermal stress on a batch of capacitors stored at high ambient temperature conditions. This provides a framework for supplementing theoretical modeling with data collected from simultaneous experiments, which is then used to validate the derived models. This work represents a first step toward combining data driven and physics-based approaches for modeling capacitor degradation. The case study shows how the proposed technique is applied to a batch of identically manufactured capacitors.

Multi-faceted modelling for strip breakage in cold rolling using machine learning

In the cold rolling process of steel strip products, strip breakage is an undesired production failure which can lead to yield loss, reduced work speed and equipment damage. To perform a root cause analysis, conventional physics-based approaches which focus on mechanical and metallurgical principles have been applied in a retrospective manner. With the advancement of data acquisition technologies, numerous process monitoring data is collected by various sensors deployed along this process; however, conventional approaches cannot take advantage of these data. In this paper, a machine learning-based approach is proposed to characterise and model strip breakage in a predictive manner. First, to match the temporal characteristic of strip breakage which occurs instantaneously, historical multivariate time-series data of a cold rolling process were extracted in a run-to-failure manner, and a sliding window strategy was adopted for data annotation. Second, breakage-centric features were identified from three facets – physics-based approaches, empirical knowledge and data-driven features. Finally, these features were used as inputs for strip breakage modelling using recurrent neural networks (RNNs), which are specialised in discovering underlying patterns embedded in time-series data. An experimental study using real-world data collected from a cold-rolled electrical steel strip manufacturer revealed the effectiveness of the proposed approach.

Optimal low-level control strategies for a high-performance hybrid electric power unit

In this paper we present models and optimization algorithms to compute the optimal low-level control strategies for hybrid electric powertrains. Specifically, we study the minimum-fuel operation of a turbocharged internal combustion engine coupled to an electrical energy recovery system, consisting of a battery and two motors connected to the turbocharger and to the wheels, respectively. First, we combine physics-based modeling approaches with neural networks to identify a piecewise affine model of the power unit accounting for the internal dynamics of the engine, and formulate the minimum-fuel control problem for a given driving cycle. Second, we parse the control problem to a mixed-integer linear program that can be solved with off-the-shelf optimization algorithms that guarantee global optimality of the solution. Finally, we validate our model against a high fidelity nonlinear simulator and showcase the presented framework with a case-study for racing applications. Our results show that cylinder deactivation and turbocharger electrification can decrease fuel consumption up to 4% and 8%, respectively.

A hybrid dynamic model for the AMBIDEX tendon-driven manipulator

Tendon-driven actuation allows for light and compact manipulator designs with enhanced safety features. One of the key challenges in model-based control of tendon-driven robots is the increased complexity of the dynamic model, due in large part to difficult-to-model behavior like nonlinear dynamic deformation and friction at the tendons. While purely data-driven modeling approaches, e.g., neural networks, free one from dealing with complex and often error-prone mechanics models, they usually do not generalize well to diverse tasks, and also do not offer the needed intuitive understanding or predictive power of traditional mechanics-based models. In this paper, we present a hybrid modeling approach for complex tendon-driven robots, which effectively complement the limitations of pure physics-based and data-driven learning-based approaches. Rigid multibody equations of motion are augmented with (i) a configuration-dependent viscous-Coulomb friction model and (ii) a recurrent neural network that captures the tendon dynamics, and estimates link joint angles from the motor positions, velocities, and torques. Experiments involving a two-dof tendon-driven parallel wrist mechanism and the 7-dof AMBIDEX tendon-driven manipulator validate the performance advantages of our hybrid model-based control framework.