3D Convolution Conjugate Gradient Inversion of Potential Fields in Acoculco Geothermal Prospect, Mexico

Potential field data have long been used in geophysical exploration for archeological, mineral, and reservoir targets. For all these targets, the increased search of highly detailed three-dimensional subsurface volumes has also promoted the recollection of high-density contrast data sets. While there are several approaches to handle these large-scale inverse problems, most of them rely on either the extensive use of high-performance computing architectures or data-model compression strategies that may sacrifice some level of model resolution. We posit that the superposition and convolutional properties of the potential fields can be easily used to compress the information needed for data inversion and also to reduce significantly redundant mathematical computations. For this, we developed a convolution-based conjugate gradient 3D inversion algorithm for the most common types of potential field data. We demonstrate the performance of the algorithm using a resolution test and a synthetic experiment. We then apply our algorithm to gravity and magnetic data for a geothermal prospect in the Acoculco caldera in Mexico. The resulting three-dimensional model meaningfully determined the distribution of the existent volcanic infill in the caldera as well as the interrelation of various intrusions in the basement of the area. We propose that these intrusive bodies play an important role either as a low-permeability host of the heated fluid or as the heat source for the potential development of an enhanced geothermal system.

Erratum to: Model compression and simplification pipelines for fast deep neural network inference in FPGAs in HEP

A Correction to this paper has been published: 10.1140/epjc/s10052-021-09770-w

Improving the Robustness of Model Compression by On-Manifold Adversarial Training

Despite the advance in deep learning technology, assuring the robustness of deep neural networks (DNNs) is challenging and necessary in safety-critical environments, including automobiles, IoT devices in smart factories, and medical devices, to name a few. Furthermore, recent developments allow us to compress DNNs to reduce the size and computational requirements of DNNs to fit them into small embedded devices. However, how robust a compressed DNN can be has not been well studied in addressing its relationship to other critical factors, such as prediction performance and model sizes. In particular, existing studies on robust model compression have been focused on the robustness against off-manifold adversarial perturbation, which does not explain how a DNN will behave against perturbations that follow the same probability distribution as the training data. This aspect is relevant for on-device AI models, which are more likely to experience perturbations due to noise from the regular data observation environment compared with off-manifold perturbations provided by an external attacker. Therefore, this paper investigates the robustness of compressed deep neural networks, focusing on the relationship between the model sizes and the prediction performance on noisy perturbations. Our experiment shows that on-manifold adversarial training can be effective in building robust classifiers, especially when the model compression rate is high.

Combustion Characteristics and NO Formation Characteristics Modeling in a Compression Ignition Engine Fuelled with Diesel Fuel and Biofuel

Compression ignition engine modeling draws great attention due to its high efficiency. However, it is still very difficult to model compression ignition engine due to its complex combustion phenomena. In this work, we perform a theoretical study of steam injection being applied into a single-cylinder four-strokes direct-injection and naturally aspirated compression ignition engine running with diesel and biodiesel fuels in order to improve the performance and reduce NO emissions by using a two-zone thermodynamic combustion model. The results obtained from biodiesel fuel are compared with the ones of diesel fuel in terms of performance, adiabatic flame temperatures, and NO emissions. The steam injection method could decrease NO emissions and improve the engine performances. The results showed that the NO formation characteristics considerably decreased and the performance significantly increased with the steam injection method. The relative errors for computed nitric oxide concentration values of biodiesel fuel and diesel fuel in comparison to the measured ones are 2.8% and 1.6%, respectively. The experimental and theoretical results observed show the highly satisfactory coincidences.