KEK Applied Research Laboratory

Improved structure–property relationships for activated carbon were obtained by devising realistic, large-scale, structural models.

Reliability-based design optimization is often a very computationally expensive process that determines the best design that satisfles a set of constraints with a specifled probability, given uncertainty in the inputs to the design. The Monte Carlo method used in this work to assess the uncertainty in a design given the input uncertainty is made computationally feasible through the use of kriging models as approximations to the original subsystem analyses. The reliability-based design optimization method described in this work uses Simulated Annealing to direct the optimization process. By using a kriging model as an approximation, additional uncertainty, namely model uncertainty, is incorporated into the design models and is included in the uncertainty assessment. During the reliabilitybased design optimization method described in this work, the number of samples used in the Monte Carlo simulation is controlled by the current temperature of the Simulated Annealing algorithm. More samples are used to improve precision as the solution nears the optimum. The method is demonstrated with the design of a satellite, and the results of not including and including the model uncertainty are presented.

An interpolating kriging model, though it will always return the observations exactly, may not provide a good representation of the computer simulation at other values within the input domain. Without access to additional and potentially costly validation observations, it is difficult to determine if a kriging model is a good representation of the original computer model. One method to determine the predictive quality of a kriging model is to use leaveone-out cross-validation. A second difficulty with creating kriging models is a lack of diagnostic tests to determine how to improve the kriging model to result in a better estimation of the original computer model. This paper presents developments of diagnostic tools for creating kriging models. A computationally efficient form for the leave-one-out cross-validation residual and the variance at the left out location is presented. The standardized residuals can then be used to test if all of the observations appear to come from the Gaussian spatial process specified by the kriging model. This lack of fit may be the result of: 1) erroneous data, 2) the form of the kriging model is not sufficient to estimate the observations as a Gaussian process, 3) or the range of the model is not well represented by a single spatial random process. Two practical examples are provided to demonstrate how to interpret the results and make decisions on how to improve the predictive capability of the kriging model. The first example is a one-dimensional adiabatic flame temperature calculation. The second problem is a two-dimensional Branin test function.

Automatic target recognition (ATR) for 3D synthetic aperture sonar (SAS) imagery is an intrinsic challenge in highly cluttered ocean environments, especially for objects partially or completely buried in the sediment. Conventional dynamic range compression (DRC) techniques such as log-compression, which is a type of tone mapping intended to appeal to the human visual system, can further obscure the sonar signatures of these already physically occluded objects and lead to suboptimal downstream ATR performance, particularly for convolutional neural networks (CNNs). In this paper, we present a novel machine learning-based approach for tone mapping sub-bottom SAS imagery as a pre-processing stage in the 3D SAS ATR pipeline. This learned tone mapping function can be jointly optimized with a CNN-based ATR algorithm. We train and validate our method on measured volumetric SAS data captured by the Sediment Volume Search Sonar (SVSS) system.

Synthetic aperture sonar (SAS) is used extensively in underwater imaging for visualizing the seafloor and objects present on it. However, processing SAS images can be time-consuming and tedious, with machine learning techniques being ineffective due to the lack of available data. In particular, automated target recognition (ATR) with 3D SAS data for machine learning is challenging in many ways due to the complexity with working with 3D volumetric data. Recently, researchers have introduced generative adversarial networks (GANs) to help perform 2D SAS image generation for data augmentation. Following this line of work in this paper, we introduce a 3D-GAN architecture to generate photorealistic 3D SAS data which matches the fidelity of real data. In particular, we discuss novel latent space sampling and normalization to help 3D GANs overcome mode collapse for generating volumetric SAS information. Experimental results are shown on real 3D SAS data, showing the potential of using 3D GANs for dataset augmentation in the future.