Online Stage Research Articles

A data-driven model reduction method for parabolic inverse source problems and its convergence analysis

In this paper, we propose a data-driven model reduction method to solve parabolic inverse source problems with uncertain data efficiently. Our method consists of offline and online stages. In the offline stage, we explore the low-dimensional structures in the solution space of parabolic partial differential equations (PDEs) in the forward problems with a given class of source functions and construct a small number of proper orthogonal decomposition (POD) basis functions to achieve significant dimension reduction. Equipped with the POD basis functions, we can solve the forward problems extremely fast in the online stage. Thus, we develop a fast algorithm to solve the optimization problem in parabolic inverse source problems, which is referred to as the POD method. Moreover, we design an a posteriori algorithm to find the optimal regularization parameter in the optimization problem using the proposed POD method without knowing the noise level. Under a weak regularity assumption on the solution of the parabolic PDEs, we prove the convergence of the POD method in solving the forward parabolic PDEs. In addition, we obtain the error estimate of the POD method for parabolic inverse source problems. Finally, we present numerical examples to demonstrate the accuracy and efficiency of the proposed method. Numerical results show that the POD method provides considerable computational savings over the finite element method while maintaining the same accuracy.

Removing the performance bottleneck of pressure–temperature flash calculations during both the online and offline stages by using physics-informed neural networks

Pressure–temperature (PT) flash calculations are a performance bottleneck of compositional-flow simulations. With the sparse grid surrogate, the computing burden of PT flash calculations is shifted from the online stage to the offline stage of the compositional-flow simulations, and a great acceleration is achieved. It is known that the data-driven neural network can also be a surrogate of PT flash calculations. However, flash calculations are carried out in the training stage, i.e., the offline stage, which means the computing burden of PT flash calculations still exists in the offline stage. With physics-informed neural networks, the two heavy-burden routines of PT flash calculations, the successive substitution technique and stability analysis, are not carried out in the offline stage, and therefore, the computing burden in the offline stage is removed. After training, the phase condition and the compositions are the output of the neural network. The numerical experiments demonstrate the correctness and the applicability of the work. To the best of our knowledge, this is the first work to remove the performance bottleneck of PT flash calculations during both the online and offline stages of compositional-flow simulations.

Determining Commercial Parking Vacancies Employing Multiple WiFiRSSI Fingerprinting Method

In this study, we implemented a parking occupancy/vacancy detection system (POVD) in a scaled-down model of a parking system for commercial centers by employing multiple WiFi access points. By exploiting the presence of WiFi routers installed in a commercial establishment, the WiFi’s received signal strength indicator (RSSI) signals were collected to establish the parking fingerprints and then later used to predict the number of occupied/vacant slots. Our extensive experiments were divided into two phases, namely: offline training and online matching phases. During the offline stage, the POVD collects available WiFi RSSI readings to determine the parking lot’s fingerprint based on a given scenario and stores them in a fingerprint database that can be updated periodically. On the other hand, the online stage predicts the number of available parking slots based on the actual scenario compared to the stored database. We utilized multiple router setups in generating WiFi signals and exhaustively considered all possible parking scenarios given the combination of 10 maximum access points and 10 cars. From two testing locations, our results showed that, given a parking area dimension of 13.40 m2 and 6.30 m2 and with the deployment of 4 and 10 routers, our system acquired the best accuracy of 88.18% and 100%, respectively. Moreover, the developed system serves as experiential evidence on how to exploit the available WiFi RSSI readings towards the realization of a smart parking system.

A Fast and Scalable Computational Framework for Large-Scale High-Dimensional Bayesian Optimal Experimental Design

We develop a fast and scalable computational framework to solve Bayesian optimal experimental design problems governed by partial differential equations (PDEs) with application to optimal sensor placement by maximizing expected information gain (EIG). Such problems are particularly challenging due to the curse of dimensionality for high-dimensional parameters and the expensive solution of large-scale PDEs. To address these challenges, we exploit two fundamental properties: (1) the low-rank structure of the Jacobian of the parameter-to-observable map, to extract the intrinsically low-dimensional data-informed subspace, and (2) a series of approximations of the EIG that reduce the number of PDE solves while retaining high correlation with the true EIG. Based on these properties, we propose an efficient offline-online decomposition for the optimization problem. The offline stage dominates the cost and entails precomputing all components that require PDE solves. The online stage optimizes sensor placement and does not require any PDE solves. For the online stage, we propose a new greedy algorithm that first places an initial set of sensors using leverage scores and then swaps the selected sensors with other candidates until certain convergence criteria are met, which we call a swapping greedy algorithm. We demonstrate the efficiency and scalability of the proposed method by both linear and nonlinear inverse problems. In particular, we show that the number of required PDE solves is small, independent of the parameter dimension, and only weakly dependent on the data dimension for both problems.

Open-Set Fault Diagnosis via Supervised Contrastive Learning With Negative Out-of-Distribution Data Augmentation

Fault diagnosis in an open world refers to the diagnosis tasks that need to cope with previously unknown faults in the online stage. It faces a great challenge yet to be addressed—that is, the online data of unknown faults may be classified as normal samples with a high probability. In this article, we develop an effective solution for this challenge by using supervised contrastive learning to learn a discriminative and compact embedding for the known normal situation and fault situations. Specifically, in addition to contrasting a given sample with other instances as is the case in conventional contrastive learning methods, our training scheme contrasts the normal samples with negative augmentations of themselves. The negative out-of-distribution data is generated by the Soft Brownian Offset sampling method to simulate the previously unknown faults. Computational experiments are conducted on the Tennessee Eastman Process benchmark dataset and a practical plasma etching process dataset. The proposed method achieves significant improvement compared with four existing methods under three open-set fault diagnosis circumstances, i.e., balanced open-set fault diagnosis, imbalanced fault diagnosis, and few-shot fault diagnosis. This demonstrates its great potentials in real world fault diagnosis applications.

Online Adaptive Algorithm for Constraint Energy Minimizing Generalized Multiscale Discontinuous Galerkin Method

In this research, we propose an online basis enrichment strategy within the framework of a recently developed constraint energy minimizing generalized multiscale discontinuous Galerkin method. Combining the technique of oversampling, one makes use of the information of the current residuals to adaptively construct basis functions in the online stage to reduce the error of multiscale approximation. A complete analysis of the method is presented, which shows the proposed online enrichment leads to a fast convergence from multiscale approximation to the fine-scale solution. The error reduction can be made sufficiently large by suitably selecting oversampling regions and the number of oversampling layers. Further, the convergence rate of the enrichment algorithm depends on a factor of exponential decay regarding the number of oversampling layers and a user-defined parameter. Numerical results are provided to demonstrate the effectiveness and efficiency of the proposed online adaptive algorithm.

A novel distributed forecasting method based on information fusion and incremental learning for streaming time series

Real-time algorithms have to adapt and adjust to new incoming patterns to provide timely and accurate responses. This paper presents a new distributed forecasting algorithm for streaming time series called StreamWNN. StreamWNN starts with an offline stage in which a forecasting model based on tuples of information fusion is created with historical data. In particular, this model consists of the fusion of patterns composed of past values of the time series with the future values of their k-nearest neighbors. Afterwards, streaming data starts to arrive. The model is incrementally updated in the online stage using a buffer with streaming data that more accurately matches the current model patterns. The model can be updated daily, monthly, quarterly or based on error thresholds. The methodology has been applied to Spanish electricity demand time series providing more accurate results when the model is updated incrementally. The best error results are obtained with the daily update of the model, resulting in an error between 2% and 3.5% depending on the prediction horizon. The model provides better error results than other algorithms.

A remaining useful life prediction method of aluminum electrolytic capacitor based on wiener process and similarity measurement

Complex characteristics such as non-linearity and multi-stage are usually presented during the degradation process of aluminum electrolytic capacitors (AECs). Therefore, it is difficult to accurately predict the capacitors’ remaining useful life (RUL), which is significant to the system’s reliability. Based on the model of Wiener process with linear drift, a simple two-stage RUL predicting method for AECs is proposed in this paper. In the offline stage, the Box–Cox transformation (BCT) is first utilized to linearize the degraded data. Then the model parameters of BCT are identified, and the Shapiro–Wilk test is employed to verify whether the transformed data meet the linearity. In the online stage, based on similarity measurement, the drift coefficient of the Wiener process is updated in real-time. By combining BCT and similarity measurement, the nonlinear and multi-stage problems are effectively solved. Finally, the proposed method is verified on NASA’s accelerated degradation data set. Experimental results show that the proposed method can automatically learn the capacitor’s complex degradation characteristics and accurately predict its RUL.

Radar Target Localization with Multipath Exploitation in Dense Clutter Environments

The performance of classic radar geometry based on the line-of-sight (LOS) signal transmitted from radar to the target in the free space is affected by multipath echoes in urban areas, where non-line-of-sight (NLOS) signals reflected by obstacles are received by the radar. Based on prior information of the urban situation, this article proposes a novel two-stage localization algorithm with multipath exploitation in a dense clutter environment. In the offline stage, multipath propagation parameters of uniformly distributed samples in the radar field of view are predicted by the ray-tracing technique. In the online stage, a rough location of the target is estimated by the maximum similarity between measurements and the predicted parameters of reference samples at different locations. The similarity is described by the likelihood between measurements and the predicted multipath parameters with respect to all possible associated hypotheses. A gating threshold is derived to exclude less likely hypotheses and reduce the computational burden. The accurate target location is acquired by a non-linear least squares (NLS) optimization of the associated multipath components. Simulation results in various noise conditions show that the proposed method provides robust and accurate target localization results under dense clutter conditions, and the offline pre-calculation of ray-tracing ensures the real-time performance of the proposed localization algorithm. The root mean square error (RMSE) of simulation results shows the advantage of the proposed method over the existing method. The presented results suggest that the proposed method can be applied to NLOS target localization applications in complex environments.

An Offline-Online Decomposition Method for Efficient Linear Bayesian Goal-Oriented Optimal Experimental Design: Application to Optimal Sensor Placement

Bayesian optimal experimental design (OED) plays an important role in minimizing model uncertainty with limited experimental data in a Bayesian framework. In many applications, rather than minimizing the uncertainty in the inference of model parameters, one seeks to minimize the uncertainty of a model-dependent quantity of interest (QoI). This is known as goal-oriented OED (GOOED). Here, we consider GOOED for linear Bayesian inverse problems governed by large-scale models represented by partial differential equations (PDE) that are computationally expensive to solve. In particular, we consider optimal sensor placement by maximizing an expected information gain (EIG) for the QoI. We develop an efficient method to solve such problems by deriving a new formulation of the goal-oriented EIG. Based on this formulation we propose an offline-online decomposition scheme that achieves significant computational reduction by computing all of the PDE-dependent quantities in an offline stage just once, and optimizing the sensor locations in an online stage without solving any PDEs. Moreover, in the offline stage we need only to compute low-rank approximations of two Hessian-related operators. The computational cost of these low-rank approximations, measured by the number of PDE solves, does not depend on the parameter or data dimensions for a large class of elliptic, parabolic, and sufficiently dissipative hyperbolic inverse problem that exhibit dimension-independent rapid spectra decay. We carry out detailed error analysis for the approximate goal-oriented EIG due to the low-rank approximations of the two operators. Furthermore, in the online stage we extend a swapping greedy method to optimize the sensor locations developed in our recent work that is demonstrated to be more efficient than a standard greedy method. We conduct a numerical experiment for a contaminant transport inverse problem with an infinite-dimensional parameter field to demonstrate the efficiency, accuracy, and both data- and parameter-dimension independence of the proposed algorithm.

Accelerating transmission-constrained unit commitment via a data-driven learning framework

As a fundamental task in power system operations, transmission-constrained unit commitment (TCUC) decides ON/OFF state (i.e., commitment) and scheduled generation for each unit. Generally, TCUC is formulated as a mixed-integer linear programming (MILP) and must be resolved within a limited time window. However, due to the NP-hard property of MILP and the increasing complexity of power systems, solving the TCUC within a limited time is computationally challenging. Regarding the computation challenge, the availability of historical TCUC data and the development of the machine learning (ML) community are potentially helpful. To this end, this paper designs an ML-aided framework that can leverage historical data in enabling computation improvement of TCUC. In the offline stage, ML models are trained to predict the commitments based on historical TCUC data. In the online stage, the commitments are quickly predicted using the well-trained ML. Furthermore, a feasibility checking process is conducted to ensure the commitment feasibility. As a result, only a reduced TCUC with fewer binary variables needs to be solved, leading to computation acceleration. Case studies on an IEEE 24-bus and a practical 5655-bus system show the effectiveness of the presented framework.

Data-Driven Reduced Order Modelling for Patient-Specific Hemodynamics of Coronary Artery Bypass Grafts with Physical and Geometrical Parameters

In this work the development of a machine learning-based Reduced Order Model (ROM) for the investigation of hemodynamics in a patient-specific configuration of Coronary Artery Bypass Graft (CABG) is proposed. The computational domain is referred to left branches of coronary arteries when a stenosis of the Left Main Coronary Artery (LMCA) occurs. The method extracts a reduced basis space from a collection of high-fidelity solutions via a Proper Orthogonal Decomposition (POD) algorithm and employs Artificial Neural Networks (ANNs) for the computation of the modal coefficients. The Full Order Model (FOM) is represented by the incompressible Navier-Stokes equations discretized using a Finite Volume (FV) technique. Both physical and geometrical parametrization are taken into account, the former one related to the inlet flow rate and the latter one related to the stenosis severity. With respect to the previous works focused on the development of a ROM framework for the evaluation of coronary artery disease, the novelties of our study include the use of the FV method in a patient-specific configuration, the use of a data-driven ROM technique and the mesh deformation strategy based on a Free Form Deformation (FFD) technique. The performance of our ROM approach is analyzed in terms of the error between full order and reduced order solutions as well as the speed-up achieved at the online stage.

An online decision-making strategy for routing of electric vehicle fleets

As environmental awareness grow, many organizations seek to implement Electric Vehicle (EV) fleets. Nonetheless, EVs’ low driving ranges and high recharging times, and the limited Charging Stations (CS) availability make their management more challenging than conventional vehicles. The Electric Vehicle Routing Problem (E-VRP) tackles these challenges. However, many E-VRP variants drop relevant operational constraints, use overly simple models, or do not address route update solutions during operation. This work introduces a strategy to compute EV routes and update them according to observed traffic scenarios. By using an event-based EV state-space model, the strategy tracks relevant variables to account for multiple realistic elements, including nonlinear recharging function, partial recharging, mass-dependent energy consumption, maximum CS capacities, and time-dependent travel times. First, an Offline E-VRP (Off-E-VRP) variant is solved to find initial route candidates. Then, routes are periodically updated during operation according to traffic and EV state measurements by solving an Online E-VRP (On-E-VRP) variant. Genetic Algorithms (GA) are implemented to solve the problems via novel encoding and genetic operators. Finally, simulation results show that the strategy enables the fleet to fulfil its delivery duties, the pre-operation stage provides adequate initial route candidates, and the online stage can improve performance and service quality.

A non-intrusive model order reduction approach for parameterized time-domain Maxwell's equations

<p style='text-indent:20px;'>We present a non-intrusive model order reduction (NIMOR) approach with an offline-online decoupling for the solution of parameterized time-domain Maxwell's equations. During the offline stage, the training parameters are chosen by using Smolyak sparse grid method with an approximation level <inline-formula><tex-math id="M1">\begin{document}$ L $\end{document}</tex-math></inline-formula> (<inline-formula><tex-math id="M2">\begin{document}$ L\geq1 $\end{document}</tex-math></inline-formula>) over a target parameterized space. For each selected parameter, the snapshot vectors are first produced by a high order discontinuous Galerkin time-domain (DGTD) solver formulated on an unstructured simplicial mesh. In order to minimize the overall computational cost in the offline stage and to improve the accuracy of the NIMOR method, a radial basis function (RBF) interpolation method is then used to construct more snapshot vectors at the sparse grid with approximation level <inline-formula><tex-math id="M3">\begin{document}$ L+1 $\end{document}</tex-math></inline-formula>, which includes the sparse grids from approximation level <inline-formula><tex-math id="M4">\begin{document}$ L $\end{document}</tex-math></inline-formula>. A nested proper orthogonal decomposition (POD) method is employed to extract time- and parameter-independent POD basis functions. By using the singular value decomposition (SVD) method, the principal components of the reduced coefficient matrices of the high-fidelity solutions onto the reduced-order subspace spaned by the POD basis functions are extracted. Moreover, a Gaussian process regression (GPR) method is proposed to approximate the dominating time- and parameter-modes of the reduced coefficient matrices. During the online stage, the reduced-order solutions for new time and parameter values can be rapidly recovered via outputs from the regression models without using the DGTD method. Numerical experiments for the scattering of plane wave by a 2-D dielectric cylinder and a multi-layer heterogeneous medium nicely illustrate the performance of the NIMOR method.</p>

React to the Worst: Lightweight and Proactive Protection of Location Privacy

This work presents a novel optimal control method for privacy protection of mobility data. Protection is based on data obfuscation, consisting in sending to the geolocated service a finely tuned fake location. The objective is twofold, keeping privacy values at an acceptable level and guaranteeing a reasonable utility loss, with a lightweight algorithm able to run on mobile devices. The proposed method consists of an offline modeling stage, based on privacy worst-case anticipation, and a fast algorithm executed online. In the offline stage, the algorithm computes the average amount of allowed utility loss necessary to maintain the privacy value of the following h steps above a given lower bound. For this purpose, the worst possible scenario over the future steps is computed and compared with the privacy function of the solution obtained by an MPC method. The online stage uses the information computed offline to solve an optimization problem whose decision variable is the location to transmit and whose objective is to maintain the privacy value above a minimal level, by avoiding large utility losses. The method is validated on an instance of a database of real records and compared with a state-of-the-art competitor.

Self-Supervised Deep Tensor Domain-Adversarial Regression Adaptation for Online Remaining Useful Life Prediction Across Machines

With deep transfer learning techniques, this paper focuses on the online remaining useful life (RUL) prediction problem across different machines, and tries to address the following concerns: 1) The effect of transfer learning decreases significantly due to considerable divergence of degradation characteristic; 2) A high computational cost is raised by re-training whole model with online data; 3) Error accumulation occurs because of lacking label information of online data. In this paper, a self-supervised deep tensor domain-adversarial regression adaptation approach is proposed. In the pre-training stage, a novel tensor domain-adversarial network, with a tensorized domain discriminator, is constructed using the offline whole-life degradation data and early fault data of the target machine. A new training algorithm with an alternating minimization scheme is then built to seek the optimal core tensor and domain-invariant feature representation. In the online stage, a new self-supervised fine-tuning strategy is designed for the target network initialized from the pre-trained network. The core tensor-formed self-supervised information, extracted from the monotonicity of online degradation process, and the pseudo-supervised information from the pre-trained network are integrated to realize fast and adaptive RUL prediction. This paper takes rolling bearing as an example, and runs both cross-conditions and cross-machines experiments on three rolling bearings datasets, i.e., IEEE PHM Challenge 2012, XJTU-SY and our test rig. The results verify the use of tensor representation can facilitate regression adversarial training, and demonstrate the proposed approach can effectively improve predictive accuracy and stability under unknown online conditions.

Adaptive Protection for Active Distribution Networks: An Approach Based on Fuses and Relays With Multiple Setting Groups

Protection schemes are essential in active distribution networks and microgrids’ reliable, efficient, and flexible operation. However, the protection of these networks presents significant challenges due to operational changes, such as variations in topology, distributed energy resources connection/disconnection, and microgrid operating modes, among others. This paper proposes an adaptive protection scheme based on overcurrent devices with several setting groups based on artificial intelligence algorithms. The developed strategy is composed of two stages. In the off-line stage, a clustering technique is employed to group the active distribution network operating scenarios exhibiting similarities. The optimal settings for the protection devices are determined for each set of scenarios. On the other hand, in the on-line stage, the protection strategy’s implementation and operation, considering the active distribution network’s existing communication system, are defined. Furthermore, the approach formulates the overcurrent relay coordination as a mixed-integer non-linear optimization problem, and as a result, the optimal setting of the overcurrent protection devices is obtained. It aims to minimize the operating time, considering the transformers’ thermal limits, fuse operating curves, and overcurrent relay settings. The solution is determined by using an Augmented Lagrangian genetic algorithm. The presented protection scheme is validated on the modified IEEE 34 node test feeder, considering the main operating scenarios of the active distribution networks, such as topology changes, distributed energy resource connection/disconnection, and microgrid operating modes (on-grid and off-grid). The results obtained and its easy implementation indicates the high potential for real-life applications.

Multimodal Fused Deep Learning Networks for Domain Specific Image Similarity Search

The exponential increase in data over the past few years, particularly in images, has led to more complex content since visual representation became the new norm. E-commerce and similar platforms maintain large image catalogues of their products. In image databases, searching and retrieving similar images is still a challenge, even though several image retrieval techniques have been proposed over the decade. Most of these techniques work well when querying general image databases. However, they often fail in domain-specific image databases, especially for datasets with low intraclass variance. This paper proposes a domain-specific image similarity search engine based on a fused deep learning network. The network is comprised of an improved object localization module, a classification module to narrow down search options and finally a feature extraction and similarity calculation module. The network features both an offline stage for indexing the dataset and an online stage for querying. The dataset used to evaluate the performance of the proposed network is a custom domain-specific dataset related to cosmetics packaging gathered from various online platforms. The proposed method addresses the intraclass variance problem with more precise object localization and the introduction of top result reranking based on object contours. Finally, quantitative and qualitative experiment results are presented, showing improved image similarity search performance.

«Democratic Tsunami» as an online stage of the Catalan independence movement

Introduction. Catalan independence is an issue that has shaken Spanish politics not even for years, but for centuries. In 2012, an overwhelming majority in parliament was received by MPs who announced the Declaration of Sovereignty. In 2017, the Catalan Parliament declared independence from Spain. In turn, the Spanish authorities arrested the initiators of this decision. Protesters took to the streets against the jailing of Catalan separatist leaders. Thus, it led to the emergence of a new protest group “Democratic Tsunami” that left a significant imprint in the struggle for the independence of the region. Research aims. To describe the “Democratic Tsunami” from the points of origins, its members, organizational structure, strategies and tactics, published content, symbols, ideology, impact and state response. Methodology. The main methods are secondary data analysis (two stages) and qualitative content analysis (649 tweets). The research period is from August 2019 to February 2020. The necessary data were collected from mass media websites and social media platforms. Results. The available data allow us to single out the communication strategies of the “Democratic Tsunami” and present their examples. The scientific article describes the operation of various online communication tools (social media, mobile application) that were used by the movement to achieve its goals. It is important to note that the Spanish authorities were not ready to confront new technologies. The studied practices can be borrowed and used in any country of the world for similar purposes and processes. Conclusions. Although the “Democratic Tsunami” lasted only six months, it was able to take the pro-independence movement to a new level of development. The main feature was the work with social media platforms and the mobile application. These movement’s tools simplified the process of mobilization and coordination between leaders and protesters. They created the mechanism to transform the Catalan independence movement into a more private and highly protected protest community.

Fast Vibration Reduction Optimization Approach for Complex Thin-Walled Shells Accelerated by Global Proper Orthogonal Decomposition Reduced-Order Model

A fast vibration reduction optimization approach accelerated by the global proper orthogonal decomposition (POD) reduced-order model (ROM) is proposed, aiming at increasing the efficiency of frequency response analysis and vibration reduction optimization of complex thin-walled shells. At the offline stage, the global POD ROM is adaptively updated using the sample configurations generated by the CV (cross validation)–Voronoi sequence sampling method. In comparison to the traditional direct sampling method, the proposed approach achieves higher global prediction accuracy. At the online stage, the fast vibration reduction optimization is performed by combining the surrogate-based efficient global optimization (EGO) method and the proposed ROM. Two representative examples are carried out to verify the effectiveness and efficiency of the proposed approach, including examples of an aerospace S-shaped curved stiffened shell and a Payload Attach Fitting. The results indicate that the proposed approach achieves high prediction accuracy and efficiency through the verification by FOM and obtains better optimization ability over the direct optimization method based on FOM.