Abstract: Generally, the present disclosure is directed to systems and methods that perform adaptive optimization with improved convergence properties. The adaptive optimization techniques described herein are useful in various optimization scenarios, including, for example, training a machine-learned model such as, for example, a neural network. In particular, according to one aspect of the present disclosure, a system implementing the adaptive optimization technique can, over a plurality of iterations, employ an adaptive effective learning rate while also ensuring that the effective learning rate is non-increasing.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for decreasing neural network inference times using softmax approximation. One of the methods includes maintaining data specifying a respective softmax weight vector for each output in a vocabulary of possible neural network outputs; receiving a neural network input; processing the neural network input using one or more initial neural network layers to generate a context vector for the neural network input; and generating an approximate score distribution over the vocabulary of possible neural network outputs for the neural network input, comprising: processing the context vector using a screening model configured to predict a proper subset of the vocabulary for the context input; and generating a respective logit for each output that is in the proper subset, comprising applying the softmax weight vector for the output to the context vector.

Abstract: The invention relates to a polar plate for a fuel cell, comprising at least one single plate with two opposing main sides, wherein the single plate has, on at least one of its main sides: elevations and free spaces arranged therebetween, which form a flow structure for an operating medium of the fuel cell; and spring projections which are arranged on the elevations and are designed to yield in case of a force effect taking place orthogonally to the main sides. Furthermore, the invention relates to a fuel cell stack with a polar plate according to the invention.

Abstract: The present invention relates to an assorted co-staging and counter stage hydro-treating process configuration scheme is disclosed for deep desulfurization and deep hydro-treating of diesel range hydrocarbons for obtaining diesel product having product sulfur less than 10 ppm and cetane number more than 51.

Abstract: The present invention is related to an integrated process for enhancing the yields of propylene and other light olefins from Fluid catalytic cracking (FCC) process in combination with Oxidative propane dehydrogenation (OPDH) where in a hydrocarbon stream from Propylene recovery section consisting of propane predominantly, is converted to high value light olefins primarily C3 and C2 olefins by catalytic oxidative dehydrogenation using carbon dioxide from FCC flue gas exiting the regenerator. Several process configurations for the conversion of C3 and C4 alkanes to their respective alkenes separately or simultaneously by integrating with FCC are provided.

Abstract: The present disclosure relates to a stacker and a method to assemble a stator core of an electric machine. Disclosed is a method for stacking a stator core of an electric machine, with the steps of providing a support adjacent to the stator frame to carry at least a sheet for the stator core, grabbing the at least a sheet with an arm arranged outside the stator core, the arm has a holding device for clamping the at least a sheet, guiding the at least a sheet inside the stator frame in a horizontal direction, and arranging the at least a sheet to compose the stator core.

Abstract: A computer-implemented method, according to one embodiment, includes: establishing a recovery cluster as an acting primary cluster in response to determining that a primary cluster has gone offline. In response to determining that the primary cluster is back online, the acting primary cluster is synched with the online primary cluster by: capturing a snapshot of the acting primary cluster, and pausing a normal queue of the acting primary cluster. The snapshot of the acting primary cluster is further compared with a most recent snapshot of the online primary cluster, and each difference therebetween is added to a priority queue of the acting primary cluster. Each of the differences in the priority queue are transmitted to the online primary cluster, followed by the entries in the normal queue in response to determining that each of the differences in the priority queue have been transmitted to the online primary cluster.

Abstract: The present invention relates to a hydrocracking reactor system and a process utilizing the same for upgrading heavy hydrocarbonaceous material to value-added products. Accordingly, an aspect of the present invention includes dispersing a liquid feedstock pre-mixed with a catalyst from top of a reactor vessel to obtain dispersed droplets having a predetermined droplet size less than 500 ?m, introducing a gaseous feed comprising primarily of hydrogen from bottom of the reactor vessel to form a continuous gaseous phase throughout a cross-section of the reactor vessel, and allowing the dispersed droplets to contact the continuous gaseous phase throughout the cross-section of the reactor vessel to form reaction effluent comprising one or more lighter product hydrocarbons. The method may further include removing at least a top portion and at least a bottom portion of the reaction effluent from the reactor vessel.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for decreasing neural network inference times using softmax approximation. One of the methods includes maintaining data specifying a respective softmax weight vector for each output in a vocabulary of possible neural network outputs; receiving a neural network input; processing the neural network input using one or more initial neural network layers to generate a context vector for the neural network input; and generating an approximate score distribution over the vocabulary of possible neural network outputs for the neural network input, comprising: processing the context vector using a screening model configured to predict a proper subset of the vocabulary for the context input; and generating a respective logit for each output that is in the proper subset, comprising applying the softmax weight vector for the output to the context vector.

Abstract: The present invention provides a catalyst composition comprising rare earth exchanged USY zeolite (REUSY); pentasil zeolite; phosphorous compound; clay, silica, alumina, and spinel to enhance the catalytic activity and selectivity for light olefins in FCC operation conditions. The present invention also provides a process for the preparation of Light olefin enhancing catalyst composition with high propylene yield and coke selectivity.

Abstract: Generally, the present disclosure is directed to systems and methods that perform adaptive optimization with improved convergence properties. The adaptive optimization techniques described herein are useful in various optimization scenarios, including, for example, training a machine-learned model such as, for example, a neural network. In particular, according to one aspect of the present disclosure, a system implementing the adaptive optimization technique can, over a plurality of iterations, employ an adaptive learning rate while also ensuring that the learning rate is non-increasing.

Abstract: The present invention relates to a process and system for complete conversion of crude oils by integrating delayed coking process, high severity catalytic cracking process and naphtha cracking processes along with olefin recovery section, aromatic recovery section and gasifier section to maximize the crude oil conversion to valuable products like light olefins, aromatics and chemicals.

Abstract: The present invention provides a process for a production of light olefins and aromatics from cracked light naphtha by selective cracking. The present invention thus provides a process for up grading cracked olefinic naphtha to high value petrochemical feed stocks. This process is based on catalytic cracking in which the catalyst activity is optimized by depositing coke for production of light olefins and aromatics. The proposed process has high flexibility and can be operated either in maximizing olefins as reflected from the PIE ratio or in maximizing aromatics (BTX) at different modes of operation depending upon the product requirement.

Abstract: The present invention relates to a process for etherification of mixed olefinic light naphtha boiling in the range of C5-90° C. cut with simultaneous minimization of unreacted methanol concentration in the product. The etherification of mixed olefinic light naphtha produces the high octane blending component which can be blended directly in the gasoline pool without any recovery of the feed oxygenates like methanol, ethanol etc. which conventionally uses energy intensive separation processes.

Abstract: The present invention relates to resid processing, particularly related to conversion of resid material with maximum recovery of lighter hydrocarbons. The invented process utilizes a novel scheme for integration of solvent de-asphalting and delayed coking processes to maximize the residue conversion to valuable products, with cleaner quality of middle distillates and fuel oil products, in comparison with other integrated solvent de-asphalting and delayed coking schemes. This process also has an additional flexibility to vary the recycle quantity, without impacting fractionator operation of the delayed coking section, which further enhances the product recovery and achieves maximum conversion of the resid feedstock, with minimum impact on liquid product properties.

Abstract: Extracting financial card information with relaxed alignment comprises a method to receive an image of a card, determine one or more edge finder zones in locations of the image, and identify lines in the one or more edge finder zones. The method further identifies one or more quadrilaterals formed by intersections of extrapolations of the identified lines, determines an aspect ratio of the one or more quadrilateral, and compares the determined aspect ratios of the quadrilateral to an expected aspect ratio. The method then identifies a quadrilateral that matches the expected aspect ratio and performs an optical character recognition algorithm on the rectified model. A similar method is performed on multiple cards in an image. The results of the analysis of each of the cards are compared to improve accuracy of the data.

Abstract: The present invention relates to the low cost high performing corrosion inhibitor composition and process for preparation thereof. The corrosion inhibitor comprises of an ester of cashew nut shell liquid (CNSL), a dimer fatty acid (DFA), a long chain amine and a solvent. The corrosion inhibitor composition prevents corrosion of internal metal surface of pipelines due to petroleum products.

Abstract: The invention relates to a polar plate assembly for a fuel cell, comprising: a polar plate which comprises, on at least one of its flat sides, elevations and free spaces arranged between them, which form a flow structure for a reactant; and a supporting structure designed to prevent or at least reduce an intrusion by a gas diffusion layer adjacent to the supporting structure and/or a membrane electrode assembly into the free spaces of the flow structure. It is provided that the polar plate assembly is designed such that the polar plate protrudes into the supporting structure. The invention furthermore relates to an individual cell with a polar plate assembly according to the invention.

Abstract: A sparse dataset is encoded using a data-driven learned sensing matrix. For example, an example method includes receiving a dataset of sparse vectors with dimension d from a requesting process, initializing an encoding matrix of dimension k×d, selecting a subset of sparse vectors from the dataset, and updating the encoding matrix via machine learning. Updating the encoding matrix includes using a linear encoder to generate an encoded vector of dimension k for each vector in the subset, the linear encoder using the encoding matrix, using a non-linear decoder to decode each of the encoded vectors, the non-linear decoder using a transpose of the encoding matrix in a projected subgradient, and adjusting the encoding matrix using back propagation. The method also includes returning an embedding of each sparse vector in the dataset of sparse vectors, the embedding being generated with the updated encoding matrix.

Abstract: The present invention relates to a method of making of hydrocarbon soluble metal composition comprising of one or more metals of group VIB of the periodic table, wherein the metal having 4+ oxidation state predominantly forms highly active metal sulfide catalyst for hydro-conversion of heavy oil feedstocks in liquid phase. More particularly, present invention relates to a hydrocarbon soluble metal composition comprising of reaction products of a metal source, a lipophilic phenolic acid, a surfactant and an organophosphorus compound. The present invention also provides a one-pot process for preparation of the hydrocarbon soluble metal composition comprising reacting a metal source, a lipophilic phenolic acid, a surfactant, an organophosphorus compound and water to obtain a reaction product and drying the reaction product to obtain the hydrocarbon soluble metal composition.