Variational Bayesian inference-based polynomial chaos expansion: Application to time-variantreliability analysis

In the time-variant systems, random variables, stochastic processes, and time parameter are regarded as the inputs of time-variant computational model. This results in an even more computationally expensive model what makes the time-variant reliability analysis a challenging task. This paper addresses the problem by presenting an active learning strategy using polynomial chaos expansion (PCE) in an augmented reliability space. We first propose a new algorithm that determines the sparse representation applying statistical threshold to determine the significant terms of the PCE model. This adaptive decision test is integrated into the variational Bayesian method, improving its accuracy and reducing convergence time. The proposed method uses a composite criterion to identify the most significant time instants and the associated training points to enrich the experimental design. By simulations, we compare the performance of the proposed method with respect to other existing time-variant reliability analysis methods.

Statistics of Extreme Events in Fluid Flows and Waves

Extreme events in fluid flows, waves, or structures interacting with them are critical for a wide range of areas, including reliability and design in engineering, as well as modeling risk of natural disasters. Such events are characterized by the coexistence of high intrinsic dimensionality, complex nonlinear dynamics, and stochasticity. These properties severely restrict the application of standard mathematical approaches, which have been successful in other areas. This review focuses on methods specifically formulated to deal with these properties and it is structured around two cases: ( a) problems where an accurate but expensive model exists and ( b) problems where a small amount of data and possibly an imperfect reduced-order model that encodes some physics about the extremes can be employed.

Wave-equation time migration

Time migration, as opposed to depth migration, suffers from two well-known shortcomings: (1)approximate equations are used for computing Green’s functions inside the imaging operator; (2) in case of lateral velocity variations, the transformation between the image ray coordinates andthe Cartesian coordinates is undefined in places where the image rays cross. We show that thefirst limitation can be removed entirely by formulating time migration through wave propagationin image-ray coordinates. The proposed approach constructs a time-migrated image without relyingon any kind of traveltime approximation by formulating an appropriate geometrically accurateacoustic wave equation in the time-migration domain. The advantage of this approach is that thepropagation velocity in image-ray coordinates does not require expensive model building and canbe approximated by quantities that are estimated in conventional time-domain processing. Synthetic and field data examples demonstrate the effectiveness of the proposed approach and show that theproposed imaging workflow leads to a significant uplift in terms of image quality and can bridge thegap between time and depth migrations. The image obtained by the proposed algorithm is correctlyfocused and mapped to depth coordinates it is comparable to the image obtained by depth migration.

Bayesian Optimization for Simulation-Based Design of Multi-Model Systems

We address the problem of simulation-based design using multiple interconnected expensive simulation models, each modeling a different subsystem. Our goal is to find the globally optimal design with minimal model evaluation costs. To our knowledge, the best existing approach is to treat the whole system as a single expensive model and apply an existing Bayesian optimization (BO) algorithm. This approach is likely inefficient due to the need to evaluate all the component models in each iteration. We propose a multi-model BO approach that dynamically and selectively evaluates one component model per iteration based on linked emulators for uncertainty quantification and the system knowledge gradient (KG) as acquisition function. Building on this, we resolve problems with constraints and feedback couplings that often occur in real complex engineering design by penalizing the objective emulator and reformulating the original problem into a decoupled one. The superior efficiency of our approach is demonstrated through solving an analytical problem and a multidisciplinary design problem of electronic packaging optimization.

Dimensionless Groups of Parameters Governing the Ice-Seabed Interaction Process

Prediction of subgouge soil deformation during an ice gouging event is a challenging design factor in Arctic subsea pipelines. An accurate assessment of ice keel–seabed interaction requires expensive model testing and large deformation finite element analysis. Proposing reliable analytical/empirical solutions needs a deep understanding of the key parameters governing the problem. In this study, dimensional analysis of subgouge soil deformations was conducted and eight dimensionless groups of parameters were identified to facilitate proposing potential new solutions. A comprehensive dataset was established for horizontal and vertical subgouge deformations in both sand and clay seabed. Using the identified dimensionless groups, linear regression (LR) models were developed to estimate the horizontal and vertical deformation. Moreover, a sensitivity analysis (SA), as well as an uncertainty analysis (UA), was carried out to identify the superior LR models and the most influential parameter group. A high range of correlation coefficient (R), Nash-Sutcliffe efficiency coefficient (NSC), and variance accounted for (VAF) along with a low range of errors was achieved for the best LR model. The results of the superior LR models were also compared with the existing empirical equations. The study showed that the shear strength parameters of the seabed soil and the ratio of gouge depth to gouge width are the governing dimensionless parameters to model the horizontal and vertical subgouge soil deformations.

Animal detections vary among commonly used camera trap models

Context Understanding how different camera trap models vary in their ability to detect animals is important to help identify which cameras to use to meet the objectives of a study. Aims To compare the efficacy of four camera trap models (representing two commonly used brands of camera, Reconyx and Scoutguard) to detect small- and medium-sized mammals and birds. Methods Four camera models were placed side by side, focused on a bait station, under field conditions, and the numbers of triggers and visits by mammals and birds were compared. Trigger=camera sensor is activated and records an image of an animal. Visit=one or a sequence of triggers containing one or more images of a species, with no interval between animal images greater than 5min. Key results The Scoutguard 530V camera recorded fewer than half of the triggers and visits by all animals that the Reconyx H600, Scoutguard 560K and Keepguard 680V cameras recorded. The latter three cameras recorded similar numbers of visits by mammals, but the Reconyx H600 recorded fewer triggers by medium-sized mammals than the Keepguard 680V. All camera models failed to detect a substantial proportion of the total known triggers and visits by animals, with a greater proportion of visits detected (14–88%) than triggers (5–83%). All camera models recorded images with no animals present (blanks), with Reconyx H600 recording the fewest blank images. Conclusions Camera trap models can vary in their ability to detect triggers and visits by small- and medium-sized mammals and birds. Some cheaper camera models can perform as well as or better than a more expensive model in detecting animals, but factors other than cost may need to be considered. Camera traps failed to detect a substantial proportion of known triggers and visits by animals. Number of visits is a more useful index of animal activity or abundance than number of triggers. Implications Variation in camera performance needs to be taken into consideration when designing or comparing camera surveys if multiple camera models are used, especially if the aim is to compare animal activity or abundance. If maximising the number of animal visits recorded at a site is important, then consideration should be given to using two or more cameras.

Quick and accurate Cellular Automata sewer simulator

As urbanisation and climate change progress, the frequency of flooding will increase. Each flood event causes damage to infrastructure and the environment. It is thus important to minimise the damage caused, which can be done through planning for events, real-time control of networks and risk management. To perform these actions, many different simulations of network behaviour are required involving complex and computationally expensive model runs. This makes fast (i.e. real-time or repetitive) simulations very difficult to carry out using traditional methods, thus there is a requirement to develop computationally efficient and accurate conceptual sewer simulators. A new Cellular Automata (CA) based sewer model is presented which is both fast and accurate. The CA model is Lagrangian in nature in that it represents the flow as blocks, and movement of the blocks through the system is simulated. To determine the number of blocks which should be moved it uses either the Manning's or Hazen–Williams equation depending on the flow conditions to calculate the permitted discharge. A case study of the sewer network in Keighley, Yorkshire, is carried out showing its performance in comparison to traditional sewer simulators. The benchmarks used to verify the results are SIPSON and SWMM5.