# US 11,915,105 B2 · Machine learning to accelerate alloy design

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,915,105 B2
- Document ID: US11915105B2
- Status: Granted
- Kind: B2
- Application number: 16/782,829
- Filed: 2020-02-05
- Published: 2020-08-13
- Granted: 2024-02-27
- Earliest claimed: 2019-02-05
- Assignees: Imagars LLC
- Applicants: Baldur Andrew Steingrimsson
- Inventors: Baldur Andrew Steingrimsson; Peter K Liaw; Xuesong Fan; Anand A Kulkarni; Duckbong Kim
- Web page: https://patentgrep.com/patent/US11915105B2

## CPC

- G06N 20/00: Machine learning
- B22F 5/04: of turbine blades
- B22F 10/366: Scanning parameters, e.g. hatch distance or scanning strategy
- B22F 10/368: Temperature or temperature gradient, e.g. temperature of the melt pool
- B22F 10/80: Data acquisition or data processing
- B22F 10/85: for controlling or regulating additive manufacturing processes
- B22F 12/45: Two or more
- B22F 12/90: Means for process control, e.g. cameras or sensors
- B29C 64/153: using layers of powder being selectively joined, e.g. by selective laser sintering or melting
- B29C 64/268: using laser beams; using electron beams [EB]
- B29C 64/393: for controlling or regulating additive manufacturing processes
- B33Y 50/02: for controlling or regulating additive manufacturing processes
- B33Y 70/00: Materials specially adapted for additive manufacturing
- C22C 19/05: with chromium
- C22C 30/00: Alloys containing less than 50% by weight of each constituent
- G06F 18/211: Selection of the most significant subset of features
- G06F 18/213: Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F 30/27: using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G16C 20/30: Prediction of properties of chemical compounds, compositions or mixtures
- G16C 20/70: Machine learning, data mining or chemometrics
- G16C 60/00: Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
- B22F 10/22: Direct deposition of molten metal
- B22F 10/25: Direct deposition of metal particles, e.g. direct metal deposition [DMD] or laser engineered net shaping [LENS]
- B22F 10/28: Powder bed fusion, e.g. selective laser melting [SLM] or electron beam melting [EBM]
- B22F 10/36: of energy beam parameters
- B33Y 10/00: Processes of additive manufacturing
- B33Y 30/00: Apparatus for additive manufacturing; Details thereof or accessories therefor
- B33Y 70/10: Composites of different types of material, e.g. mixtures of ceramics and polymers or mixtures of metals and biomaterials
- C22C 14/00: Alloys based on titanium
- C22C 16/00: Alloys based on zirconium
- C22C 21/00: Alloys based on aluminium
- C22C 27/02: Alloys based on vanadium, niobium, or tantalum
- C22C 27/025: {alloys based on vanadium}
- C22C 27/04: Alloys based on tungsten or molybdenum
- C22C 27/06: Alloys based on chromium
- G06F 2113/10: Additive manufacturing, e.g. three-dimensional [3D] printing
- G06F 2119/08: Thermal analysis or thermal optimisation
- G06N 3/02: Neural networks
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- Y02P 10/25: Process efficiency
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- B22F: WORKING METALLIC POWDER; MANUFACTURE OF ARTICLES FROM METALLIC POWDER; MAKING METALLIC POWDER (making alloys by powder metallurgy C22C); APPARATUS OR DEVICES SPECIALLY ADAPTED FOR METALLIC POWDER
- B29C: SHAPING OR JOINING OF PLASTICS; SHAPING OF MATERIAL IN A PLASTIC STATE, NOT OTHERWISE PROVIDED FOR; AFTER-TREATMENT OF THE SHAPED PRODUCTS, e.g. REPAIRING (making preforms B29B11/00; making laminated products by combining previously unconnected layers which become one product whose layers will remain together B32B37/00 - B32B41/00)
- B33Y: ADDITIVE MANUFACTURING, i.e. MANUFACTURING OF THREE-DIMENSIONAL [3D] OBJECTS BY ADDITIVE DEPOSITION, ADDITIVE AGGLOMERATION OR ADDITIVE LAYERING, e.g. BY 3D PRINTING, STEREOLITHOGRAPHY OR SELECTIVE LASER SINTERING
- C22C: ALLOYS (treatment of alloys C21D, C22F)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G16C: COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- Y02P: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS

## Abstract

This invention presents an innovative framework for the application of machine learning for identification of alloys or composites with desired properties of interest. For each output property of interest, we identify the corresponding driving (input) factors. These input factors may include the material composition, heat treatment, process, microstructure, temperature, strain rate, environment or testing mode. Our framework assumes selection of optimization technique suitable for the application at hand and data available, starting with simple linear, or quadratic, regression analysis. We present a physics-based model for predicting the ultimate tensile strength, a model that accounts for physical dependencies, and factors in the underlying physics as a priori information. In case an artificial neural network is deemed suitable, we suggest employing custom kernel functions consistent with the underlying physics, for the purpose of attaining tighter coupling, better prediction, and extracting the most out of the—usually limited—input data available.

## Claims

### Claim 1

An apparatus for predictive analytics, an apparatus that employs a prediction module, for the purpose of efficiently searching composition space of alloys or composites of interest, and hence for accelerating design or manufacturing of alloys or composites with desired material characteristics, an apparatus comprising of
a database importing module, for ingesting materials or manufacturing data,
an optional data base abstraction module, also referred to as prediction logic, for abstracting the interface between the prediction module and the database,
an optional preprocessing module, for unified comparison of materials or manufacturing data across data sets,
an optional feature selection module, for extracting features that best describe the material or manufacturing data of interest,
a prediction module, also referred to as a prediction engine, for predicting material properties or manufacturing parameters of interest, along with a corresponding composition or composition range, given the materials or manufacturing data ingested, where the prediction engine employs a prediction technique, where the prediction engine combines physics-based models, or analytical models, specific to the alloys or composites of interest, with traditional black-box prediction models, for improved prediction accuracy, and where the prediction technique employed can involve regression analysis, with relatively few unknown model parameters, when limited data is available, but a machine learning predictor, with relatively many unknown model parameters, when sufficiently large data set is available, where the prediction technique is selected such that at least one data point is available for each unknown prediction model parameter, and
an optional reporting and verification module, for reporting, evaluating or verifying the materials properties or manufacturing parameters estimated,
an apparatus presented either in the form of an integrated or an embedded application.

### Claim 2 (depends on 1)

An apparatus according to claim 1, where the data importing module employs an Export, Transform and Load operation for importing data from a source into the destination prediction database logic, when the destination prediction database logic system represents the data differently from the source.

### Claim 3 (depends on 1)

An apparatus according to claim 1, where the preprocessing module applies normalization for standardizing the input data ingested, for unified comparison, and where the standardization can involve division of endurance limit with tensile strength, in case of prediction of fatigue behavior of the alloys or composites of interest.

### Claim 4 (depends on 1)

An apparatus according to claim 1, where the feature selection module utilizes features related to composition or micro-structure of the alloys or composite of interest, process, heat treatment, temperature, environment, testing mode or loading rate.

### Claim 5 (depends on 1)

An apparatus according to claim 1, where the prediction engine is capable of accommodating models capturing physical dependencies, referred to as physics-based models, as a priori information, and correspondingly constructing dependencies within the prediction model, for purpose of expediting training of the prediction model or improving prediction accuracy.

### Claim 6 (depends on 5)

An apparatus according to claim 5, where the physics-based models employ thermodynamics, first principle, empirical rules, mesoscale models, or in case of prediction of fatigue life, models for dislocation dynamics or slip band information.

### Claim 7 (depends on 1)

An apparatus according to claim 1, where the prediction engine consists of optional “forward” prediction along with optional “inverse” prediction, where the “forward” prediction is responsible for predicting the material properties or manufacturing parameters of interest from source characteristics, where the “inverse” prediction is responsible for inferring source characteristics, in particular alloy composition, from source characteristics from material properties desired, and where the source characteristics can consist of composition, microstructure, process, heat treatment, temperature, environment, testing mode or loading rate.

### Claim 8 (depends on 1)

An apparatus according to claim 1, where the feature selection module employs canonical component analysis or regression, for purpose of selecting features capable of distinguishing between calcium-magnesium-alumino-silicate and calcium sulfate hot corrosion attacks, in air or sea water, in order to develop coatings resistant to calcium-magnesium-alumino-silicate and calcium sulfate hot corrosion.

### Claim 9 (depends on 1)

An apparatus according to claim 1, where the prediction engine utilizes a parametrized model, referred to as augmented Statistical Fatigue Life model, a model capable of accounting for multiple sources impacting fatigue life of additively manufactured components, for purpose of accurately predicting the fatigue life.

### Claim 10 (depends on 1)

An apparatus according to claim 1, where the prediction engine supports inverse design representations.

### Claim 11 (depends on 1)

An apparatus according to claim 1, where the prediction engine can employ custom kernel functions consistent with underlying physics, for the purpose of attaining tighter coupling, better prediction, than with generic, non-custom kernel functions, and for extracting the most out of the materials or manufacturing data ingested.

### Claim 12

A method for predictive analytics, one that incorporates a prediction step, for the purpose of efficiently searching composition space of alloys or composites of interest, and hence for accelerating design or manufacturing of alloys or composites with desired material characteristics, a method utilizing
a database importing step, for ingesting materials or manufacturing data,
an optional preprocessing step, for unified comparison of materials or manufacturing data across data sets,
an optional feature selection step, for extracting the features primarily impacting the material properties of interest,
a prediction step, for predicting material properties or manufacturing parameters of interest, along with a corresponding composition or composition range, given the materials or manufacturing data ingested, where the prediction step employs a prediction technique, where the prediction step combines physics-based models, or analytical models, specific to alloys or composites of interest, with traditional black-box prediction models, for improved prediction accuracy, and where the prediction technique employed can involve regression analysis, with relatively few unknown model parameters, when limited data is available, but a machine learning predictor, with relatively many unknown model parameters, when a sufficiently large data set is available, where the prediction technique is selected such that at least one data point is available for each unknown prediction model parameter, and
a reporting and evaluation step, for reporting or evaluating the materials properties or manufacturing parameters of interest.

### Claim 13 (depends on 12)

A method according to claim 12, where the data importing step employs an Export, Transform and Load operation for importing data from a source into the destination predictive analytics system, when the destination system represents the data differently from the source.

### Claim 14 (depends on 12)

A method according to claim 12, where the preprocessing step applies a normalization step for standardizing the input data, and where the standardization involves division of endurance limit with tensile strength, in case of prediction of fatigue behavior of alloys or composites of interest.

### Claim 15 (depends on 12)

A method according to claim 12, where the feature selection step utilizes features related to composition or microstructure of the alloy or composite of interest, process, heat treatment, temperature, environment, testing mode or loading rate.

### Claim 16 (depends on 12)

A method according to claim 12, where the prediction step is capable of accommodating models capturing physical dependencies, referred to as physics-based models, as a priori information, and correspondingly constructing dependencies within the prediction model, for the purpose of expediting training or improving prediction accuracy.

### Claim 17 (depends on 16)

A method according to claim 16, where the physics-based models involve thermodynamics, first principle, empirical rules or mesoscale models, or in case of prediction of fatigue life, models for dislocation dynamics or slip band information.

### Claim 18 (depends on 12)

A method according to claim 12, where the prediction step consists of an optional “forward” prediction step along with an optional “inverse” prediction step, where the “forward” prediction step is responsible for predicting the material properties or manufacturing parameters of interest from source characteristics, where the “inverse” prediction step is responsible for inferring source characteristics, in particular alloy composition, from source characteristics from material properties targeted, and where the source characteristics can consist of composition, microstructure, process, heat treatment, temperature, environment, testing mode or loading rate.

### Claim 19 (depends on 12)

A method according to claim 12, where the feature selection step employs canonical component analysis or regression analysis, for purpose of selecting features capable of distinguishing between calcium-magnesium-alumino-silicate and calcium sulfate hot corrosion attacks, with or without influence of sea salt, in order to develop coatings resistant to calcium-magnesium-alumino-silicate and calcium sulfate hot corrosion.

### Claim 20 (depends on 16)

A method according to claim 16, where the physics-based models involve thermodynamics, first principle, empirical rules or mesoscale models, or in case of prediction of fatigue life, models for dislocation dynamics or slip band information.

### Claim 21 (depends on 12)

A method according to claim 12, where the prediction step supports inverse design representations.