Bp Corporation North America Inc.

National Technology & Engineering Solutions of Sandia, LLC

Air Products and Chemicals, Inc.

Scott Roberts

Richard Wilsak

Dean B Comstock

Ronald D Stefanski

Charles J Snider

Matthew David Smith

Kevin Matthew Potter

John P Korbin

Carianne Martinez

Emily Donahue

Lincoln Collins

Jeffrey T Miller

George A Huff

Ruth Ann Doyle

Jeffrey Amelse

Giorgio Zoia

Joe Moore

Richard A Wilsak

Kevin Ferguson

Brian Slusar

Kathryn Whitaker

Stephen Islas

Jim Puissant

Rose M Janulis

Steven Norwood

Chyau Lin

Chris Schneller

Jeffrey Allen Amelse

Ronald Stefanski

Digital image segmentation is provided. The method comprises training a neural network for image segmentation with a labeled training dataset from a first domain, wherein a subset of nodes in the neural net are dropped out during training. The neural network receives image data from a second, different domain. A vector of N values that sum to 1 is calculated for each image element, wherein each value represents an image segmentation class. A label is assigned to each image element according to the class with the highest value in the vector. Multiple inferences are performed with active dropout layers for each image element, and an uncertainty value is generated for each image element. Uncertainty is resolved according to expected characteristics. The label of any image element with an uncertainty above a threshold is replaced with a new label corresponding to a segmentation class based on domain knowledge.

Uncertainty-refined image segmentation under domain shift

A method for digital image segmentation is provided. The method comprises training a neural network for image segmentation with a labeled training dataset from a first domain, wherein a subset of nodes in the neural net are dropped out during training. The neural network receives image data from a second, different domain. A vector of N values that sum to 1 is calculated for each image element, wherein each value represents an image segmentation class. A label is assigned to each image element according to the class with the highest value in the vector. Multiple inferences are performed with active dropout layers for each image element, and an uncertainty value is generated for each image element. The label of any image element with an uncertainty value above a predefined threshold is replaced with a new label corresponding to the class with the next highest value.

Processes for recovering paraxylene from at least two feedstreams containing xylene isomers. The process includes directing to a paraxylene recovery zone comprising at least one crystallization zone, a paraxylene-lean feedstream having a paraxylene to total xylene isomer ratio of 0.50 or less, and also directing to the paraxylene recovery zone, a paraxylene-rich feedstream having a paraxylene to total xylene isomer ratio of greater than 0.50. A paraxylene-containing product stream is recovered from the paraxylene recovery zone having a paraxylene to total xylene isomer ratio greater than that of the paraxylene-rich feedstream. The process provides improvements in paraxylene recovery efficiency and/or cost effectiveness.

Process for recovering paraxylene from at least two feedstreams containing xylene isomers

Wafer carrier washing and drying apparatus and method, especially useful for the semiconducting industry.

Fixture drying apparatus and method

A filter column apparatus comprising a filtration zone and a reslurry zone. These zones are separated by a barrier wall or are in substantial cooperation with each other. Also disclosed is a process for separating at least a portion of at least one substantially solid component from a solid-liquid stream comprising the substantially solid component and at least one substantially liquid component. Also disclosed is a process for forming a substantially solids containing packed bed. Also disclosed is a process for purifying paraxylene in a filtration zone.

Apparatus and process for the separation of solids and liquids

Processes for separating a solid from a solids-liquids slurry under conditions that prevent ingress of molecular oxygen are advantageous where the presence of molecular oxygen would otherwise reduce efficiencies, contribute to limited product yields, and potentially compromise safe operation of the process or downstream unit operations. Among the various embodiments disclosed herein, is a process utilizing filter columns as solid-liquid separators in combination with crystallization and reslurry unit operations to recover a product component from an initial feed mixture of miscible components. Embodiments of the disclosed processes may include the separation and purification of a product component using a crystallizer in series with a filter column followed by a chemical reactor, using a reslurry drum in series with a filter column, and using a combination of crystallizers and/or reslurry drums in series with at least one filter column. The use of filter columns in such processes is particularly attractive because these columns, when operated under the conditions disclosed herein, limit the undesirable ingress of molecular oxygen into the various process streams, and further improve process efficiencies related to both product component recovery and operating costs.

Melt-crystallization separation and purification process

A process for separating solids from liquids in a filtration zone defined by a higher concentration zone and a lower concentration zone separated by a filter. The process includes the steps of directing a slurry feed comprising a liquid and solids into the higher concentration zone, directing a displacement fluid to the higher concentration zone and passing at least a portion of the liquid through a filter to the filtrate zone, producing a filtrate.

Solid-liquid separation process

A pressure swing adsorption process to separate para-xylene and ethylbenzene from C aromatics which uses a para-selective, non-acidic, medium pore molecular sieve of the MFI structure type and is operated isothermally in the vapor phase at elevated temperatures and pressures is integrated with crystallization to produce para-xylene product. A fixed bed of adsorbent is saturated with pX and EB, which are preferentially adsorbed, the feed is stopped, and lowering the partial pressure desorbs the pX and EB. The process effluent, which is rich in pX and EB, is crystallized to obtain para-xylene product.

Para-xylene production process integrating pressure swing adsorption and crystallization

Growing community of inventors

Naperville, IL, United States of America

Scott Roberts