AI INTEGRATED SYSTEM, METHOD, AND COMPUTER READABLE MEDIUM FOR CHARACTERIZING MICROSPHERICAL MATERIALS - Patent application

Abstract

Disclosed are methods, systems, and computer readable media for providing a microsphere analysis tool (MAT), comprising the steps of: (a) receiving image data representing microscopic image data, spectroscopic image data, or a combination of microscopic and spectroscopic image data of a microsphere material, (b) dividing the image data into tiled image regions, (c) stitching the tiled image regions together to form a composite image, (d) classifying, with a trained machine learning model, each tiled image region as corresponding to one of: (i) a particle, (ii) an intraparticle pore, or (iii) an interparticle void, and (e) characterizing at least one of: (i) a composition, (ii) a composition-specific particle size distribution, (iii) a crystalline phase, (iv) a degree of crystallinity, or (v) a crystal-specific particle size distribution of the microsphere material.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Patent Application No. 63 / 269,034, filed March 8, 2022, the disclosure of which is incorporated herein by reference in its entirety.

[0002] The present disclosure relates to systems, methods, and computer-readable media for acquiring microscopic and spectroscopic image data of microspherical materials, classifying the image data of the microspherical materials, and characterizing compositional and structural properties of the microspherical materials based on the image data. [Background technology]

[0003] Cracking is a process used to break down complex organic molecules, such as high molecular weight hydrocarbons, into simpler molecules by inducing the cleavage of carbon-carbon bonds, which typically occurs in the presence of a catalyst. The rate of cracking and the nature of the end products depend on the conditions under which the process is carried out, such as temperature, pressure, the choice of chemical feedstock, and the properties of the catalyst and catalyst support material.

[0004] Catalytic cracking processes involve the presence of acid catalysts (usually solid acids such as alumina, silica-alumina, and zeolites) that promote the heterolytic cleavage of bonds to generate ion pairs with opposite charges, usually a carbocation and a highly unstable hydride anion. Both carbon-localized free radicals and cations are highly unstable and undergo chain rearrangement, CC scission at the β-position (i.e. cracking), and intra- and intermolecular hydrogen or hydride transfer processes. In both types of processes, the corresponding reactive intermediates (radicals, ions) are perpetually regenerated, so that the reaction proceeds by a self-propagating chain mechanism. The reaction chain is eventually terminated by the recombination of the radicals or ions.

[0005] One widely used cracking process is fluid catalytic cracking (FCC), which typically uses microspherical catalyst particles suspended in an increasing stream of feed hydrocarbon to form a fluidized bed. Typical FCC catalysts contain active crystalline aluminosilicates, primarily faujasite-type zeolites (e.g., various forms of Y zeolite, such as REY, HY, REHY, USY, REUSY, etc.), to improve the yield of liquefied petroleum gas (LPG), LPG olefins such as propylene and butylene, and / or light cycle oil (LCO).

[0006] In some instances, the zeolite and active inorganic oxide matrix components used as active cracking components are compounded in granules with a binder, typically formed from an amorphous gel or sol such as a silica sol, that serves to bind the components together when dried. The binder itself may or may not be active. A filler, such as a kaolin type clay, makes up the remainder of the catalyst composition. The oxide matrix material may be attrition resistant, selective with respect to the production of hydrocarbons without excessive coke formation, and not easily deactivated by metals. A cracking active matrix (active matrix) may be formed to support the overall cracking reaction.

[0007] In other instances, the zeolite itself may function as a binder and be the only binder material present. Such FCC catalysts are described in "Continuous Age Distribution Method for Catalytic Cracking. 1. Proof of Principle" by DM Stockwell, Ind. Eng. Chem. Res. 2015, 54, pp. 5921-593. In the in-situ process, the active matrix, which may be calcined kaolin clay or alumina, is mixed with kaolin clay and spray-dried to produce microspheres. Metakaolin is produced from kaolinite by calcination. Y zeolite is crystallized with metakaolin to dissolve silica from the calcined clay. If necessary, additional silica, caustic, and seeds are added to the crystallization reactor. The Y zeolite in the obtained in-situ crystallized microspheres also serves as an acid catalyst to bind the microspheres together. Incorporated FCC catalysts and additives such as ZSM-5 additives often contain kaolin clay as a filler. The in-situ microspheres do not contain kaolin clay as a filler since the kaolin originally present becomes metakaolin after calcination of the microspheres, and the metakaolin is a reactant for producing Y zeolite. Summary of the Invention [Problem to be solved by the invention]

[0008] Pores are typically classified according to size as follows: (i) pores with a width greater than about 50 nm (0.05 pm) are called macropores, (ii) pores with a width between 2 nm and 50 nm are called mesopores, and (iii) pores with a width less than about 2 nm are called micropores. K.S.W. Singh et al., "Reporting Physisorption Data for Gas / Solid Systems with Special Reference to the Determination of Surface Area and Porosity (Recommendations 1984)," Pure & Appl. Chem., 1985, Vol. 57, No. 4, pp. 603-619. Mercury porosity is often used to characterize macroporosity in FCC and other catalysts. Pores with a diameter greater than 2 μm (20,000 Å) are typically considered interparticle pores. However, mercury porosimetry cannot distinguish between larger intraparticle macropores and interparticle spaces.

[0009] Additionally, monitoring and characterizing FCC catalysts is necessary to better understand catalytic activity and product selectivity. Various properties of FCC catalysts, such as porosity, attrition resistance, elemental composition, crystalline phase, particle size distribution, and tolerance to impurities such as nickel, vanadium, copper, and iron, all play a role in catalytic activity and product selectivity. Accurately characterizing an FCC catalyst for a given industrial or power generation application is difficult and time-consuming. This makes analyzing and optimizing the performance of FCC catalysts extremely challenging. [Means for solving the problem]

[0010] The present disclosure presents a Microsphere Analysis Tool (MAT) that provides tools and methods for data analysis and mining of composite images of microsphere materials with trained machine learning models. More specifically, the MAT performs a method that includes the steps of (a) receiving image data representing microscopic image data, spectroscopic image data, or a combination of microscopic and spectroscopic image data of a microsphere material, (b) dividing the image data into tiled image regions, (c) stitching the tiled image regions together to form a composite image, (d) classifying, with a trained machine learning model, each tiled image region as corresponding to one of (i) a particle, (ii) an intra-particle pore, or (iii) an inter-particle void, and (e) characterizing at least one of (i) the composition, (ii) the composition-specific particle size distribution, (iii) the crystalline phase, (iv) the degree of crystallinity, or (v) the crystal-specific particle size distribution of the microsphere material.

[0011] In at least one embodiment, the tiled image data is acquired by scanning electron microscopy (SEM).

[0012] In at least one embodiment, the tiled image data is obtained by cross-sectional scanning electron microscopy.

[0013] In at least one embodiment, the tiled image data is acquired by backscattered electron scanning electron microscopy (BSE-SEM).

[0014] In at least one embodiment, the tiled image data is obtained by transmission electron spectroscopy (TEM).

[0015] In at least one embodiment, the tiled image data is acquired by cross-sectional transmission electron spectroscopy.

[0016] In at least one embodiment, the tiled image data is acquired by Energy Dispersive X-ray Spectroscopy (EDX).

[0017] In at least one embodiment, the tiled image data is two-dimensional image data.

[0018] In at least one embodiment, the tiled image data is three-dimensional image data.

[0019] In at least one embodiment, the tiled image data is obtained by a partially automated process.

[0020] In at least one embodiment, the tiled image data is acquired by a fully automated process.

[0021] In at least one embodiment, the tiled image data is obtained by an unsupervised process.

[0022] In at least one embodiment, the tiled image data is obtained by dividing a grid into a row-column pattern. In at least one embodiment, the tiled image data includes overlapping image tiles. In at least one embodiment, the overlapping image tiles overlap by between 5% and 20%.

[0023] In at least one embodiment, the tiled image data is stitched together using a trained machine learning model to form a composite image.

[0024] In at least one embodiment, the tiled image data is acquired at a magnification between 100x and 6,000x.

[0025] In at least one embodiment, the tiled image data is acquired at a magnification between 100x and 1,000x.

[0026] In at least one embodiment, the tiled image data is acquired at a magnification between 250x and 550x.

[0027] In at least one embodiment, the tiled image data is acquired at a magnification between 1,000x and 6,000x.

[0028] In at least one embodiment, the tiled image data is acquired at a magnification between 4,000x and 5,500x.

[0029] In at least one embodiment, the tiled image data is acquired at a magnification between 20,000x and 500,000x.

[0030] In at least one embodiment, the tiled image data includes a microscopic image and a spectroscopic image acquired at substantially the same magnification.

[0031] In at least one embodiment, the tiled image data includes a microscopic image and a spectroscopic image acquired of substantially the same field of view.

[0032] In at least one embodiment, the tiled image data includes a microscopic image and a spectroscopic image acquired at substantially the same magnification and substantially the same field of view.

[0033] In at least one embodiment, one or more regions of the composite image are classified as corresponding to at least one of a particle, an intra-particle pore, or an inter-particle void.

[0034] In at least one embodiment, one or more regions of the composite image are classified to correspond to at least one of clay, alumina, or pores. In at least one embodiment, the microsphere material comprises at least one of alumina or clay-based silica-alumina. In at least one embodiment, the alumina comprises at least one of gibbsite, flash-calcined gibbsite, bayerite, boehmite, or alumina. In at least one embodiment, the clay-based silica-alumina comprises at least one of kaolin clay, hydrous clay, calcined kaolin clay, or sodium silicate.

[0035] In at least one embodiment, the particle classified region is further classified to include at least one of potassium, cesium, calcium, barium, strontium, copper, yttrium, phosphorus, sulfur, selenium, fluorine, chlorine, bromine, iodine, lanthanum, cerium, aluminum, silicon, sodium, carbon, oxygen, iron, vanadium, or nickel.

[0036] In at least one embodiment, the particle classified region is further classified as at least one of iron, vanadium, and nickel.

[0037] In at least one embodiment, the microspherical material is nodular.

[0038] In at least one embodiment, the microsphere material is a fluid catalytic cracking (FCC) catalyst.

[0039] In at least one embodiment, the FCC catalyst is an equilibrium catalyst (Ecat).

[0040] In at least one embodiment, the alumina particle crystallinity of the microsphere material is between 20% and 50% boehmite single crystal.

[0041] In at least one embodiment, the alumina particle crystallinity of the microsphere material is between 20% and 30% boehmite single crystal.

[0042] In at least one embodiment, the alumina particle crystallinity of the microsphere material is between 35% and 45% boehmite single crystal.

[0043] In at least one embodiment, the microsphere material is embedded in an epoxy resin prior to imaging.

[0044] In at least one embodiment, the epoxy resin is polished and carbon coated prior to imaging.

[0045] In at least one embodiment, a machine learning model is trained to distinguish between particles, intra-particle pores, and inter-particle voids using manually annotated image data as a ground truth reference.

[0046] In at least one embodiment, the machine learning model includes an artificial neural network.

[0047] In at least one embodiment, the machine learning model includes a convolutional neural network.

[0048] In at least one embodiment, the machine learning model includes a recurrent neural network.

[0049] The present disclosure also provides systems and computer readable media for performing the disclosed methods.

[0050] The accompanying drawings, which form a part of this specification, illustrate several embodiments and, together with the description, serve to explain the principles disclosed herein. The patent or application file contains at least one color drawing. A copy of the color drawing(s) will be provided by the Office upon request and payment of the necessary fee with this patent or patent application publication. [Brief description of the drawings]

[0051] [Figure 1] 1 illustrates an exemplary system 1000 for performing a compliant method. [Diagram 2] FIG. 20 is a flow diagram illustrating a method 2000 for training a machine learning model. [Diagram 3] 3 is a flow diagram showing further details of the method 2000 of FIG. 2. [Figure 4] FIG. 3 is a flow diagram illustrating a method 3,000 for classifying one or more properties of a microsphere material. [Diagram 5] 3A-3C show example images generated in steps 3100, 3300, 3400, and 3500 of method 3000. [Figure 6]Figure 6A shows a composite BSE-SEM image stitched together with the color-coded composition predictions of sample A by MAT. Figure 6B shows a tiled BSE-SEM image of the color-coded composition predictions of sample A by MAT. Figure 6C shows a tiled EDX ground truth image of Figure 6B. [Figure 7] FIG. 13 shows a composite image of stitched image tiles of color-coded composition predictions for sample A by MAT. [Figure 8] 1 shows a plot of the alumina particle size distribution for sample A. [Figure 9] Figure 9A shows a stitched composite BSE-SEM image of the color-coded composition prediction of sample B by MAT. Figure 9B shows a tiled BSE-SEM image of the color-coded composition prediction of sample B by MAT. Figure 9C shows a tiled EDX ground truth image of Figure 9B. [Figure 10] 1 shows a plot of the alumina particle size distribution for sample B. [Figure 11A] Figure 1 shows a composite BSE-SEM image of the color-coded composition predictions of sample C by MAT. [Figure 11B] FIG. 13 shows a composite BSE-SEM image of the color-coded composition predictions of sample D by MAT. [Figure 11C] FIG. 13 shows a merged composite BSE-SEM image of color-coded composition predictions for sample E by MAT. [Figure 11D] FIG. 13 shows a composite BSE-SEM image of the color-coded composition predictions of sample F by MAT. [Figure 12] A plot of impurity iron nodules is shown. [Figure 13] Figure 1 shows tiled BSE-SEM images of color-coded composition predictions for sample C by MAT. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0052] In this disclosure, the term "catalyst" or "catalyst composition" or "catalyst material" includes, for example, a material that promotes a reaction. For example, an "oxidizing material" promotes an oxidation reaction with one or more components of the exhaust gas.

[0053] The term "microsphere" includes, for example, particles having a diameter of about 1 μm to about 500 μm. Microspheres need not be uniform in size and may be solid or hollow. Microspheres may comprise a matrix having pores.

[0054] The term "pore" includes, for example, an opening or depression in the surface of a microsphere, or a tunnel in the microsphere. A pore may be a single tunnel or may be connected to other tunnels in a continuous network throughout the microsphere. Pores are characterized based on porosity, pore size, pore volume, surface area, density, pore size distribution, and pore length.

[0055] The term "matrix" includes the solid materials that form the FCC microspheres, for example. The matrix may refer to clay (such as calcined kaolin clay) and / or alumina.

[0056] The term "interparticle voids" includes, for example, voids within a material, such as voids between microspheres.

[0057] The term "extra-large macropores" includes, for example, macropores having an equivalent diameter greater than 1 micron.

[0058] The term "large macropores" includes, for example, macropores larger than 0.5 microns.

[0059] The term "zeolite" includes crystalline materials, which may be particulate, for example, in combination with one or more promoter metals used as catalysts. Zeolites are materials based on an extended three-dimensional network of oxygen ions with approximately tetrahedral sites and a substantially uniform pore distribution, with substantially uniform pore sizes of 20 Å or less. Zeolites include, for example, aluminosilicates, aluminophosphates, ferrisilicates, and silica aluminophosphates. The term "aluminophosphate" refers to specific examples of molecular sieves that contain aluminum and phosphate atoms. Aluminophosphates are crystalline materials with rather uniform pore sizes. The listed examples are not necessarily exhaustive of all possible zeolite materials.

[0060] The term "about" or "approximately" allows for a 5% deviation from a given value. Alternatively, the term "about" includes the standard deviation or variance, if any, for a given value.

[0061] The term "no additives" includes that the material contains less than 5% of the target element. In some embodiments, the material may have less than 3% of the target element.

[0062] The systems and methods described herein may be implemented on a wide range of microsphere materials. In some implementations, for example, the microsphere material may be an FCC catalyst. The catalyst may include, for example, a zeolite, a clay, and a binder with or without phosphorus or alumina. In some embodiments, the binder is selected from colloidal silica, peptized alumina, aluminum phosphate, and combinations thereof. In some embodiments, the microspheres include clay and a zeolite. In other implementations, the microsphere material may be a material that has not yet been crystallized in-situ. In some embodiments, the microsphere material may be nodular. In some embodiments, the FCC catalyst may be an equilibrium catalyst (Ecat).

[0063] In some embodiments, the microsphere material comprises at least one of alumina or clay-based silica-alumina. In some embodiments, the alumina comprises at least one of gibbsite, bayerite, flash calcined gibbsite, boehmite, or alumina. In some embodiments, the clay-based silica-alumina comprises at least one of kaolin clay, hydrous clay, calcined kaolin clay, or sodium silicate. In some embodiments, the particle classified region is further classified to include at least one of aluminum, silicon, sodium, carbon, oxygen, iron, vanadium, or nickel. In some embodiments, the compositions of the present disclosure comprise microspheres comprising macropores having an equivalent diameter greater than 1 micron. In some embodiments, the equivalent diameter of the macropores is in the range of about 1 micron to about 30 microns, e.g., about 1 micron to about 10 microns, about 1 micron to about 20 microns, about 1 micron to about 30 microns, about 2 microns to about 30 microns, about 2 microns to about 10 microns, about 2 microns to about 20 microns, about 2 microns to about 30 microns, about 5 microns to about 10 microns, about 5 microns to about 15 microns, about 5 microns to about 20 microns, about 5 microns to about 25 microns, about 5 microns to about 30 microns, about 10 microns to about 12 microns, about 10 microns to about 14 microns, about 10 microns to about 16 microns, about 10 microns to about 18 microns, about 10 microns to about 20 microns, about 10 microns to about 22 microns, about 10 microns to about 24 microns, about 10 microns to about 26 microns, about 10 microns to about 28 microns, about 10 microns to about 30 microns, etc.

[0064] In some embodiments, the compositions of the present disclosure include extra-large macropores that are greater than 10 microns. In some embodiments, a portion of the macropores that are greater than 10 microns may be in the range of about 50 ppm to about 1850 ppm, e.g., about 100 ppm to about 1850 ppm, about 150 ppm to about 1850 ppm, about 200 ppm to about 1850 ppm, about 250 ppm to about 1850 ppm, about 300 ppm to about 1850 ppm, about 350 ppm to about 1850 ppm, about 400 ppm to about 1850 ppm, or about 500 ppm to about 1850 ppm. pm, about 450ppm to about 1850ppm, about 500ppm to about 1850ppm, about 600ppm to about 1850ppm, about 650ppm to about 1850ppm, about 700ppm to about 1850ppm etc., about 750ppm to about 1850ppm, about 800ppm to about 1850ppm, about 850ppm to about 1850ppm, about 900ppm to about 1850ppm, about 950ppm to about 1850ppm, Approximately 1,000ppm to approximately 1850ppm, approximately 1050ppm to approximately 1850ppm, approximately 1100ppm to approximately 1850ppm, approximately 100ppm to approximately 1850ppm, approximately 100ppm to approximately 1850ppm m, about 1150ppm to about 1850ppm, about 1200ppm to about 1850ppm, about 1250ppm to about 1850ppm, about 1300ppm to about 1850ppm, about 1350ppm to about 185 0 ppm, about 1400 ppm to about 1850 ppm, about 1450 ppm to about 1850 ppm, about 1500 ppm to about 1850 ppm, about 1550 ppm to about 1850 ppm, about 1600 ppm to about 1850 ppm, about 1650 ppm to about 1850 ppm, about 1700 ppm to about 1850 ppm, about 1750 ppm to about 1850 ppm, about 1800 ppm to about 1850 ppm, etc.

[0065] In some embodiments, the compositions of the present disclosure comprise microspheres comprising extra-large macropores characterized by an average pore area of ​​about 0.2 μm. 2 ~about 2.0μm 2 range, for example, about 0.3 μm 2 ~about 2.0μm 2 , about 0.4μm 2 ~about 2.0μm 2 , about 0.5 μm2 ~about 2.0μm 2 , about 0.6 μm 2 ~about 2.0μm 2 , about 0.7μm 2 ~about 2.0μm 2 , about 0.8μm 2 ~about 2.0μm 2 , about 0.9 μm 2 ~about 2.0μm 2 , about 1.0μm 2 ~about 2.0μm 2 , about 1.1μm 2 ~about 2.0μm 2 , about 1.2μm 2 ~about 2.0μm 2 , about 1.3μm 2 ~about 2.0μm 2 , about 1.4μm 2 ~about 2.0μm 2 , about 1.5 μm 2 ~about 2.0μm 2 , about 1.6 μm 2 ~about 2.0μm 2 , about 1.7μm 2 ~about 2.0μm 2 , about 1.8μm 2 ~about 2.0μm 2 , about 1.9 μm 2 ~about 2.0μm 2 etc. in the range.

[0066] In some embodiments, the compositions of the present disclosure include microspheres comprising extra-large macropores characterized by an average porosity, in some embodiments, the average porosity of the macropores is in the range of about 5% to about 21%, e.g., about 6% to about 21%, about 7% to about 21%, about 8% to about 21%, about 9% to about 21%, about 10% to about 21%, about 11% to about 21%, about 12% to about 21%, about 13% to about 21%, about 14% to about 21%, about 15% to about 21%, about 16% to about 21%, about 17% to about 21%, about 18% to about 21%, about 19% to about 21%, or about 20% to about 21%.

[0067] In some embodiments, the compositions of the present disclosure comprise 90In some embodiments, the microspheres include microspheres that include extra-large macropores characterized by the d 90 is in the range of about 0.5 to about 1.1 μm, for example, 0.6 to about 1.1 μm, 0.7 to about 1.1 μm, 0.8 to about 1.1 μm, 0.9 to about 1.1 μm, 1.0 to about 1.1 μm, about 0.5 to about 0.7 μm, 0.5 to about 0.9 μm, 0.6 to about 0.8 μm, 0.8 to about 1.0 μm, etc.

[0068] In some embodiments, the compositions of the present disclosure include microspheres comprising single boehmite crystals and kaolin clay. In some embodiments, the alumina particles of the microspheres have a crystallinity of single boehmite crystals in the range of about 20% to about 50%, such as about 20% to about 30%, about 35% to about 45%, etc.

[0069] In some embodiments, the compositions of the present disclosure include single boehmite crystals, aggregated boehmite crystals, and microspheres comprising kaolin clay. 50 is in the range of about 70 nm to about 500 nm, for example, in the range of about 70 nm to about 100 nm, about 80 nm to about 90 nm, about 300 nm to about 500 nm, about 350 nm to about 450 nm, etc.

[0070] Aluminosilicates may have an open three-dimensional framework structure consisting of corner-sharing TO4 tetrahedra, where T is Al or Si, or optionally P. Anionic framework charge-balancing cations are loosely bound to framework oxygens, and the remaining pore volume is filled with water molecules. Non-framework cations are generally exchangeable, and water molecules are generally removable.

[0071] Any type of molecular sieve, such as ABW, ACO, AEI, AEL, AEN, AET, AFG, AFI, AFN, AFO, AFR, AFS, AFT, AFX, AFY, AHT, ANA, APC, APD, AST, ASV, ATN, ATO, ATS, ATT, ATV, AWO, AWW, BCT, BEA, BEC, BIK, BOG, BPH, BRE, CAN, CAS, SCO, CFI, SGF, CGS, CHA, CHI, CLO, CON, CZP, DAC, DDR, DFO, DFT, DOH, DON, EAB, EDI, EMT, EON, EPI, ERI, ESV, ETR, EUO, FAU, FER, FRA, GIS, GIU, GME, GON, GOO, HEU, IFR, IHW, ISV, ITE, ITH, ITW, IWR, IWW, JBW, KFI, LAU, LEV, LIO, LIT, LOS, LOV, LTA, LTL, LTN, MAR, MAZ, MEI, MEL, MEP, MER, MFI, MFS, MON, MOR, MOZ, MSO, MTF, MTN, MTT, MTW, MWW, NAB, NAT, NES, NON, NPO, NSI, OBW, OFF, OSI, OSO, OWE, PAR, PAU, PHI, PON, RHO, RON, RRO, RSN, RTE, RTH, RUT, RWR, RWY, SAO, SAS, SAT, SAV, SBE, SBS, SBT, SFE, SFF, SFG, SFH, SFN, SFO, SGT, SOD, SOS, SSY, STF, STI, STT, TER, THO, TON, TSC, UEI, UFI, UOZ, USI, UTL, VET, VFI, VNI, VSV, WIE, WEN, YUG, ZON, etc., or combinations thereof can be used.

[0072] In some embodiments, the zeolite is selected from the group including faujasite (Y, USY, REO-Y), ZSM-5, Beta, FER, ZSM-11, ZSM-22, ZSM-35, MSM-68, chabazite, and combinations thereof. In some embodiments, the zeolite is modified with an element. In at least one embodiment, the element is selected from the group including phosphorus and transition metals (scandium, titanium, vanadium, chromium, manganese, iron, cobalt, nickel, copper, zinc, yttrium, zirconium, niobium, molybdenum, technetium, ruthenium, rhodium, palladium, silver, cadmium, hafnium, tantalum, tungsten, rhenium, osmium, iridium, platinum, gold, mercury).

[0073] Zeolites may also include compositions of aluminosilicates, borosilicates, gallosilicates, MeAPSO, and MeAPO, for example. The zeolite may be a natural or synthetic zeolite, for example, faujasite, chabazite, clinoptilolite, mordenite, silicalite, zeolite X, zeolite Y, ultrastable zeolite Y, ZSM-5, ZSM-12, SSZ-3, SAPO 5, offretite, or beta zeolite. In one or more specific embodiments, the zeolite is selected from type A, chabazite, erionite, ZSM-5, ZSM-11, ZSM-23, ZSM-48, ferrierite, stilbite, faujasite, mordenite, type L, omega, beta, AlPO4, borosilicate, MeAPO, MeAPSO, and SAPO. In some embodiments, the molecular sieve has a BEA structure.

[0074] The term "binder" includes, for example, compounds that promote adhesion and / or cohesion between particles of the zeolite material to be bound beyond the physical adsorption that may exist without the binder. Binders may be, for example, colloidal silica, peptized alumina, aluminum phosphate, and combinations thereof.

[0075] system FIG. 1 illustrates an example system 1000 for performing methods according to the present disclosure.

[0076] 1, the system 1000 may be implemented using a client / server architecture including one or more client processing devices 11001-N for executing user applications and one or more server processing devices 12001-N for executing server applications. The client processing device 1100 may communicate with the server processing device 1200 via an electronic interface 1300, e.g., a wide area network (WAN) interface, a local area network (LAN) interface, or a wired and / or wireless communication interface, such as the Internet. Alternatively, the system 1000 may be implemented as a stand-alone processing device, e.g., the processing device 11001.

[0077] The client side processing devices 1100 may be implemented as thin clients or thick clients using, for example, a personal computer, a server terminal, a mobile device, etc., and may be in the form of, for example, a desktop, laptop, or handheld device. As shown in FIG. 1, the client side processing devices 1100 may each include one or more processing units 1110 and memory 1120 operably coupled by a bus 1130.

[0078] The processing unit 1110 may include one or more processors (e.g., microprocessors) and associated hardware, software, and / or hardwired logic circuitry programmed to execute methods according to the present disclosure. The processors may operate alone or in parallel. The memory 1120 may include non-transitory computer-readable media, such as both read-only memory (ROM) and random access memory (RAM). At various times, computer-readable instructions, data structures, program modules, and data necessary for execution of the methods disclosed herein may be stored in the ROM and / or RAM portions of the memory 1120. In particular, the memory 1120 may store an operating system, one or more client-side application programs (e.g., computer or mobile application programs) and / or program modules, and program data. The bus 1130 may include a memory bus or memory controller, a peripheral bus, and a local bus, each of which may be implemented using any of a variety of bus architectures.

[0079] Each of the client processing devices 1100 may also include one or more user input devices 1140 and output devices 1150. The output devices may include, for example, a monitor, display, speakers, and / or printers that output information to a user. The user input devices 1140 may include, for example, a keyboard, microphone, scanner, and / or pointing device such as a mouse or touch screen for inputting commands or data in cooperation with a graphical user interface displayed on the display or monitor.

[0080] The server side processing devices 1200 may be implemented using a personal computer, a network server, a web server, a file server, etc. As shown in FIG. 1, the server side processing devices 1200 may each include one or more processing units 1210 and memory 1220 operably coupled by a bus 1230.

[0081] The processing unit 1210 may include one or more processors (e.g., microprocessors) and associated hardware, software, and / or hardwired logic circuitry programmed to execute methods according to the present disclosure. The processors may operate alone or in parallel. The memory 1220 may include non-transitory computer-readable media, such as both read-only memory (ROM) and random access memory (RAM). At various times, computer-readable instructions, data structures, program modules, and data necessary for execution of the methods disclosed herein may be stored in the ROM and / or RAM portions of the memory 1220. In particular, the memory 1220 may store an operating system, one or more server-side application programs and / or program modules, and program data. The bus 1230 may include a memory bus or memory controller, a peripheral bus, and a local bus, each implemented using any of a variety of bus architectures.

[0082] In some implementations, system 1000 may further include one or more sensor inputs 1400 that provide data necessary to perform methods consistent with the present disclosure. The sensor inputs may include laboratory and / or testing equipment for collecting such data, such as a transmission electron microscope (TEM) 1410, a scanning electron microscope (SEM) 1420, and / or an energy dispersive X-ray spectrometer (EDX) 1430. The sensor inputs 1400 may also include inputs for use with related TEM, SEM, and EDX techniques, such as a cross-sectional TEM, cross-sectional SEM, or backscattered electron scanning electron microscope (BSE-SEM).

[0083] Training a machine learning model 2 is a flow diagram that generally illustrates a method 2000 for training a machine learning model to classify one or more regions and characterize one or more of the composition, composition-specific particle size distribution, or crystalline phase distribution, crystallinity, or crystal-specific particle size distribution of the particle-classified regions of the microsphere material. The method 2000 may be performed by executing computer readable instructions, data structures, and / or program modules stored in memory 1120 and / or 1220.

[0084] In step 2100, a machine learning model may be trained using a data set collected from a first set of microscopy and / or spectroscopic images. For example, data may be collected using one or more sensor inputs, such as a TEM 1410, a SEM 1420, and / or an EDX 1430. The data set may be stored in one or more of the memories 1120 and / or 1220 and may further be distributed across multiple such memories. The machine learning module may be trained and executed using one or more server-side processing units 1200, one or more client-side processing units 1100, or a combination of such units operating in series and / or parallel.

[0085] In step 2200, microscopy and / or spectroscopic image data is acquired from a new sample to be analyzed. For example, the data may be collected using the same or different inputs as in step 2100, such as the TEM 1410, SEM 1420, and / or EDX 1430.

[0086] In step 2300, the trained machine learning model is used to predict a classification of the image region and one or more characterizations of the composition, composition-specific particle size distribution, or crystal phase distribution, crystallinity, or crystal-specific particle size distribution of the particle-classified region of the microsphere material. For example, the analysis may be determined by the trained machine learning model operating on one or more server-side processing devices 1200, one or more client-side processing devices 1100, or a combination of such devices operating in series and / or parallel to output to a user using an output device 1150, such as a display or printer.

[0087] In step 2400, classification and characterization data from new samples is added to the dataset for future training of the machine learning model.

[0088] FIG. 3 is a flow diagram illustrating the method 2000 of FIG. 2 in more detail. In particular, FIG. 3 illustrates certain sub-steps that may be performed within steps 2100-2400. As shown in FIG. 3, a machine learning model is trained using a set of data collected from a first set of TEM, SEM, and / or EDX images (step 2100). TEM, SEM, and / or EDX image data of a new sample is acquired (sub-step 2210). The images may be pre-processed (e.g., cropped), and machine learning techniques are applied to these images for semantic segmentation (sub-step 2220). Individual pixels are classified according to a set of predefined key features during training and are uniquely colored based on these key features. Adjacent pixels of the same feature are grouped as objects that can be extracted for measurement (sub-steps 2310-2340). Statistical analysis is performed with measurements of all objects in all feature classes (sub-step 2350). Data collected from the new sample may then be added to the data set for training future machine learning models (step 2400).

[0089] In step 2200, the system 1000 may extract training data including one or more TEM, SEM, and / or EDX images and identify a parameter or feature from each image such that each parameter or feature is collected from each image in the multiple samples. For example, the system may extract a set of training data as described above with respect to FIGS.

[0090] In step 2300, the training data may be classified according to at least one structural, fundamental, or physical characteristic described in the training data. For example, the training data may be classified according to the identification of pores or interparticle voids, or the detection of alkali metals (such as, but not limited to, sodium, potassium, cesium, etc.), alkaline earth metals (such as, but not limited to, calcium, barium, strontium, etc.), transition metals (such as, but not limited to, iron, nickel, copper, yttrium, and vanadium), post-transition metals (such as, but not limited to, aluminum), metalloids (such as, but not limited to, silicon), non-metallic elements (such as, phosphorus, sulfur, selenium, halogens, etc.), and actinides and lanthanides (such as, but not limited to, lanthanum, cerium, etc.).

[0091] In step 2300, the system can generate a machine learning model that may be trained based on the selected features or parameters to predict compositional and / or structural properties of the FCC catalyst. The machine learning model can be trained using supervised or unsupervised training techniques.

[0092] Supervised learning allows predictions based on data models that can be generated from a training set. Suitable supervised learning techniques may include, for example, deep learning, and suitable supervised learning techniques may include, for example, artificial neural networks, convolutional neural networks, recurrent neural networks, support vector machines, linear discriminant analysis, and logistic regression techniques. Other suitable supervised and unsupervised learning techniques may be used, such as Bayesian net genetic algorithms / genetic programming, simulated annealing, entanglement hierarchies of sets, recursive partitioning, clustering, hidden Markov models, fuzzy methods, semantic networks, naive Bayes similarity mapping, support vector machines, self-organizing maps, and Gaussian processing techniques. In some embodiments, supervised and unsupervised learning models may be trained using manually annotated image data as a ground truth standard to distinguish between particles, intra-particle pores, and inter-particle voids.

[0093] Microsphere Analysis Tool (MAT) 4 is a flow diagram illustrating a method 3000 for acquiring and characterizing microspheric material based on image data. As shown in FIG. 3, microscopic and spectroscopic image data is acquired in multiple image tiles (step 3100). The tiled image data may be two-dimensional or three-dimensional image data acquired at substantially the same accelerating voltage, substantially the same magnification level and working distance or field of view, and with substantially the same dwell time. In some embodiments, the tiles may be acquired at a higher magnification, thereby increasing the resolution within a particular field of view.

[0094] In some embodiments, the microsphere material can be embedded in an epoxy resin that can be polished and carbon coated prior to imaging. Once the sample is positioned, the tiled image data can be acquired by a fully or partially automated process, or an unsupervised process. For example, the sample and / or sensor can be configured to move relative to one another to collect multiple image tiles acquired by grid division in a row-column pattern.

[0095] In some embodiments, the tiled image data is acquired at a fixed magnification. In some embodiments, the fixed magnification is in the range of about 100X to about 6,000X, for example, about 100X to about 1,000X, about 250X to about 550X, about 1,000X to about 6,000X, or about 4,000X to about 5,500X. In some embodiments, the fixed magnification is in the range of about 20,000X to about 500,000X, for example, about 20,000X to about 50,000X, about 20,000X to about 40,000X, about 20,000X to about 30,000X, about 100,000X to about 500,000X, about 250,000X to about 500,000X, or about 400,000X to about 500,000X.

[0096] To minimize unimaged sample area, tiles overlap adjacent tiles in two directions (horizontally and vertically). For example, image tiles may overlap horizontally and vertically by about 5%-20%, about 10%-20%, or about 15%-20%. The horizontal and vertical overlap amounts may be the same or different.

[0097] Adjacent tiles are stitched together to form a mosaic or composite image (step 3200) to avoid double counting of overlapping regions in statistical analysis. For example, tiles can be stitched together using the Microscopy Image Stitching Tool (MIST) application developed by the National Institute of Standards and Technology (NIST) for the ImageJ / Fiji imaging software suite to reduce processing time and errors in mosaic images.

[0098] Regions of the composite image may then be segmented and classified using a trained machine learning model (step 3300). For example, the machine learning model may have been trained using techniques described in U.S. Application No. 63 / 203217, incorporated herein by reference in its entirety. In some embodiments, these regions may be classified as either particles, intra-particle pores, or inter-particle voids. The particle-classified regions may then be further characterized according to one or more of composition, composition-specific particle size distribution, crystalline phase, crystallinity, or crystal-specific particle size distribution using the trained machine learning model (step 3500). Measurements of the classified or characterized features of interest may then be collected and statistically analyzed (step 3600).

[0099] Figure 5 is a flow diagram showing exemplary images generated during steps of method 3000 for classifying regions to characterize compositional and / or structural properties of microsphere materials. In particular, Figure 5 shows exemplary images for sub-steps 3100, 3300, 3400, and 3500, where BSE-SEM and EDX images were acquired with the same field of view and magnification. EXAMPLES

[0100] The following non-limiting examples are disclosed to aid in the understanding of the present disclosure and should not be construed as limiting the embodiments described herein in any way. As would be expected by one skilled in the art, variations of such embodiments, including the substitution of all equivalents now known or developed in the future, and slight changes in formulation or experimental design are intended to be within the scope of the embodiments described herein. The following non-limiting examples further illustrate various aspects and features of the disclosed systems and methods and computer-readable media.

[0101] Example 1 SEM and EDX images of all specimens of samples A and B were acquired at 5,000x magnification using the autonomous microscope and spectroscopic image acquisition functions in an unsupervised overnight session.

[0102] Samples A and B were prepared by first embedding in epoxy resin followed by polishing and carbon coating. The compositions of microspheres Samples A and B were prepared according to Table 1. Sample A microspheres contained 6% boehmite with the remainder being low-iron kaolin clay. Sample B microspheres contained 6% boehmite with the remainder being low-iron kaolin clay. Mass fractions were reported as free of volatiles.

[0103] [Table 1]

[0104] SEM images were collected at a high pixel resolution of 2,000 × 1,500 pixels, and iterative training was performed for 500 epochs to improve the accuracy of the model in distinguishing between particles, intraparticle pores, and interparticle voids.

[0105] The trained artificial intelligence model was then deployed for automated semantic segmentation of four SEM images from each sample, A and B. Once the accuracy of the segmentation was verified, recipes and macros were incorporated into the MIPAR environment to measure and quantify the composition and structural attributes of the microsphere material.

[0106] FIG. 6A is a 300× composite image of 266 tiled SEM images of sample A acquired at 5,000× magnification in an SEM during an unsupervised overnight session. Complementary EDX images acquired at the same field of view and magnification were acquired to train the machine learning model. FIG. 6B shows the alumina and clay particles identified by the machine learning model in sample A and compared to FIG. 6C, a ground truth image of sample A. In FIG. 6B, the alumina regions are colored yellow, the clay regions are colored blue, the intraparticle pores are colored green, and the interparticle voids are colored black. In FIG. 6C, the alumina regions of the EDX ground truth are colored yellow, the clay regions are colored gray, and both the intraparticle pores and the interparticle voids are colored black. The intraparticle pores and the interparticle voids may be further differentiated using different colors based on their classification as intraparticle pores and interparticle voids by the trained machine learning model, as shown in FIG. 6B.

[0107] Figure 7 shows a composite image of sample A, generated by stitching together adjacent image tiles. The crystal size distribution of the alumina particles of sample A is shown in Figure 8. The size distribution of the 26,047 alumina particles was identified as the sum of two log-normal distributions colored blue. The first log-normal distribution of sample A, colored red, is 40% boehmite single crystals and d 50 is equal to 89 nm. The second log-normal distribution of sample A, colored green, is 60% aggregates and d 50 is equal to 333 nm.

[0108] FIG. 9A shows a 300x composite image of 210 tiled SEM images of sample B acquired at 5,000x magnification by SEM in an unsupervised overnight session. FIG. 9B shows the alumina and clay particles of sample B, which was compared to FIG. 9C, the ground truth image of sample B. In FIG. 9B, the alumina regions are colored yellow, the clay regions are colored blue, the intraparticle pores are colored green, and the interparticle voids are colored black. In FIG. 9C, the alumina regions of the EDX ground truth are colored yellow, the clay regions are colored gray, and the intraparticle pores and interparticle voids are both colored black, making them indistinguishable.

[0109] The crystal size distribution of the alumina particles of sample B is shown in Figure 10. The particle size distribution of 33,162 alumina particles was identified as the sum of two log-normal distributions colored blue. The first log-normal distribution of sample B colored red is 25% boehmite single crystals, d 50 is equal to 83 nm. The second log-normal distribution of sample B, colored green, is 75% aggregated and d 50 is equal to 444 nm.

[0110] The crystal size distributions of alumina particles for samples A and B plotted in Figures 8 and 10 were further compared with the crystal size estimated from X-ray diffraction (not shown). By considering the width of the 120 reflection from the XRD diffractogram in terms of the Scherrer equation, the crystal size of samples A and B was estimated to be in the range of 80 nm to 100 nm. In comparing the XRD estimated crystal size of samples A and B with the SEM crystal size distribution determined by the machine learning model, only a small proportion of alumina particles were accurately captured by the XRD analysis (40% of sample A and 25% of sample B). In both samples A and B, the machine learning model showed a second distribution that was not reflected by the conventional characterization method.

[0111] Example 2 SEM and EDX images of whole specimens of samples C, D, E, and F were acquired at 5,000x using partially autonomous microscopy and spectroscopic image acquisition (Figs. 11A-D, Table 2). Each of Figs. 11A-D is a composite image stitched together from overlapping image tiles. Fig. 11A was stitched together from 320 image tiles, Fig. 11B from 220 image tiles, Fig. 11C from 300 image tiles, and Fig. 11D from 268 image tiles. The trained machine learning model characterized the image regions, and also the particle-classified regions, according to their composition. Specifically, the particle-classified regions characterized as iron nodule particles were colored red and quantified according to the percentage of particles with nodular impurity iron. The results of the quantification by the machine learning model are shown in Fig. 12. Specifically, Fig. 12 shows that as the wt% of impurity iron increases, the machine learning model recognized an increasing percentage of particles that partially contain iron nodules. This was consistent with the trend seen in Figure 11A-D, where the qualitative proportion of red particles increased as a function of impurity iron (wt%). Despite the increasing impurity iron wt%, these samples C, D, E, and F in separate analyses showed comparable catalytic activity, suggesting that the impurity iron wt% inferred from the analysis is not a direct or indirect indicator of catalytic performance. Figure 13 shows the individual image tiles of sample E classified and characterized according to the trained machine learning model.

[0112] SEM images were collected at a high pixel resolution of 2,000 × 1,500 pixels and tiled into 9 subframes for training. 300 epochs of training iterations were performed to refine the model's ability to distinguish between particles, intraparticle pores, and interparticle voids.

[0113] [Table 2]

[0114] The above description has been presented for purposes of illustration. It is not exhaustive or limited to the precise forms or embodiments disclosed. Modifications and adaptations of the embodiments will become apparent from review of the specification and practice of the disclosed implementations. For example, while certain elements of system 1000 have been described as being coupled together, such elements may be integrated or distributed in any suitable manner.

[0115] Moreover, although exemplary embodiments have been described herein, the scope includes any embodiment having equivalent elements, modifications, omissions, combinations (e.g., aspects across various implementations), adaptations, or alterations based on the present disclosure. Claim elements are to be interpreted broadly based on the language used in the claims, and are not limited to the examples described in this specification or in the practice of this application, which examples should be interpreted as non-exclusive. Furthermore, the steps of the disclosed methods can be modified in any manner, including rearranging the order of steps, inserting or deleting steps. In particular, independent steps may be performed in any order or in parallel.

[0116] Other embodiments will be apparent from consideration of the specification and practice of the disclosed implementations. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosed embodiments being indicated by the following claims.

Claims

1. 1. A method for characterizing a microsphere material, comprising: acquiring tiled image data representing a microscopic image of the microsphere material, a spectroscopic image of the microsphere material, or a combination of a microscopic image and a spectroscopic image of the microsphere material; stitching the tiled image data into a composite image; classifying one or more regions of the composite image with a trained machine learning model; The trained machine learning model classifies one or more classified regions of the composite image according to the following characteristics: composition, composition-specific particle size distribution, crystalline phase, Crystallinity, or Grain size distribution The method includes the step of characterizing according to at least one of the following:

2. The method of claim 1 , wherein the tiled image data is acquired by a scanning electron microscope (SEM).

3. The method of claim 1 , wherein the tiled image data is acquired by cross-sectional scanning electron microscopy.

4. The method of claim 1 , wherein the tiled image data is acquired by backscattered electron scanning electron microscopy (BSE-SEM).

5. The method of claim 1 , wherein the tiled image data is acquired by transmission electron spectroscopy (TEM).

6. The method of claim 1 , wherein the tiled image data is acquired by cross-sectional transmission electron spectroscopy.

7. The method of claim 1 , wherein the tiled image data is acquired by Energy Dispersive X-ray Spectroscopy (EDX).

8. The method of claim 1 , wherein the tiled image data is two-dimensional image data.

9. The method of claim 1 , wherein the tiled image data is three-dimensional image data.

10. The method of claim 1 , wherein the tiled image data is obtained by a partially automated process.

11. The method of claim 1 , wherein the tiled image data is obtained by a fully automated process.

12. The method of claim 1 , wherein the tiled image data is obtained by an unsupervised process.

13. The method of claim 1 , wherein the tiled image data is obtained by grid division in a row-column pattern.

14. The method of claim 13 , wherein the tiled image data comprises overlapping image tiles.

15. The method of claim 14, wherein the partially overlapping image tiles overlap by between 5% and 20%.

16. The method of claim 1 , wherein the tiled image data is stitched into a composite image by a trained machine learning model.

17. The method of claim 1 , wherein the tiled image data is acquired at a magnification between 100x and 6,000x.

18. The method of claim 1 , wherein the tiled image data is acquired at a magnification between 100x and 1,000x.

19. The method of claim 1 , wherein the tiled image data is acquired at a magnification between 250x and 550x.

20. The method of claim 1 , wherein the tiled image data is acquired at a magnification between 1,000x and 6,000x.

21. The method of claim 1 , wherein the tiled image data is acquired at a magnification of between 4,000x and 5,500x.

22. The method of claim 1 , wherein the tiled image data is acquired at a magnification between 20,000x and 500,000x.

23. The method of claim 1 , wherein the tiled image data comprises microscopic and spectroscopic images acquired at substantially the same magnification.

24. The method of claim 1 , wherein the tiled image data comprises a microscopic image and a spectroscopic image acquired of substantially the same field of view.

25. The method of claim 1 , wherein the tiled image data comprises microscopic and spectroscopic images acquired at substantially the same magnification and substantially the same field of view.

26. The method of claim 1 , wherein the one or more regions of the composite image are classified as corresponding to at least one of a particle, an intra-particle pore, or an inter-particle void.

27. The method of claim 1 , wherein the one or more regions of the composite image are classified as corresponding to at least one of clay, alumina, and pores.

28. 28. The method of claim 27, wherein the microsphere material comprises at least one of alumina or clay-based silica-alumina.

29. 30. The method of claim 28, wherein the alumina comprises at least one of gibbsite, flash-calcined gibbsite, bayerite, boehmite, or alumina.

30. 29. The method of claim 28, wherein the clay-based silica-alumina comprises at least one of kaolin clay, hydrous clay, calcined kaolin clay, or sodium silicate.

31. 10. The method of claim 1, wherein the particle classified region is further classified as containing at least one of potassium, cesium, calcium, barium, strontium, copper, yttrium, phosphorus, sulfur, selenium, fluorine, chlorine, bromine, iodine, lanthanum, cerium, aluminum, silicon, sodium, carbon, oxygen, iron, vanadium, or nickel.

32. The method of claim 1 , wherein the particle classified region is further classified as at least one of iron, vanadium, or nickel.

33. The method of claim 1 , wherein the microspherical material is nodular.

34. The method of claim 1 , wherein the microsphere material is a fluid catalytic cracking (FCC) catalyst.

35. 35. The method of claim 34, wherein the FCC catalyst is an equilibrium catalyst (Ecat).

36. The method of claim 1, wherein the microsphere material has an alumina particle crystallinity of 20% to 50% boehmite single crystal.

37. The method of claim 1, wherein the microsphere material has an alumina particle crystallinity of 20% to 30% boehmite single crystal.

38. The method of claim 1, wherein the microsphere material has an alumina particle crystallinity of 35% to 45% boehmite single crystal.

39. The method of claim 1 , wherein the microsphere material is embedded in an epoxy resin prior to imaging.

40. 40. The method of claim 39, wherein the epoxy resin is polished and carbon coated prior to imaging.

41. 10. The method of claim 1, wherein the machine learning model is trained to distinguish between particles, intra-particle pores, and inter-particle voids using manually annotated image data as a ground truth standard.

42. The method of claim 1 , wherein the machine learning model comprises an artificial neural network.

43. The method of claim 1 , wherein the machine learning model comprises a convolutional neural network.

44. The method of claim 1 , wherein the machine learning model comprises a recurrent neural network.

45. 1. A system including at least one processor, The processor, obtaining tiled image data representing a microscopic image of the microspherical material, a spectroscopic image of the microspherical material, or a combination of a microscopic image and a spectroscopic image of the microspherical material; stitching the tiled image data together to form a composite image; classifying one or more regions of the composite image with a trained machine learning model; The trained machine learning model classifies one or more classified regions of the composite image according to the following characteristics: composition, composition-specific particle size distribution, crystalline phase, Crystallinity, or Grain size distribution 4. A system configured to characterize a signal according to at least one of:

46. A system according to claim 45, configured to carry out the method according to any one of claims 1 to 44.

47. When executed by at least one processor, the method causes the at least one processor to: obtaining tiled image data representing a microscopic image of the microspherical material, a spectroscopic image of the microspherical material, or a combination of a microscopic image and a spectroscopic image of the microspherical material; stitching the tiled image data together to form a composite image; classifying one or more regions of the composite image with a trained machine learning model; The trained machine learning model classifies one or more classified regions of the composite image according to the following characteristics: composition, composition-specific particle size distribution, crystalline phase, Crystallinity, or Grain size distribution A non-transitory computer-readable medium having encoded thereon instructions for causing a device to be characterized according to at least one of the following:

48. 48. The non-transitory computer readable medium of claim 47, wherein the instructions further cause the at least one processor to perform the method of any one of claims 1 to 44.

Images