# US 12,002,552 B2 · Systems and methods for prediction of polymer properties > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,002,552 B2 - Document ID: US12002552B2 - Status: Granted - Kind: B2 - Application number: 17/604,621 - Filed: 2020-04-16 - Published: 2022-02-10 - Granted: 2024-06-04 - Earliest claimed: 2019-04-19 - Assignees: Georgia Tech Research Corporation - Applicants: Georgia Tech Research Corporation - Inventors: Rampi Ramprasad; Anand Chandrasekaran; Chiho Kim - Web page: https://patentgrep.com/patent/US12002552B2 ## CPC - 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 - G06N 3/045: Combinations of networks - G06N 3/048: Activation functions - G06N 3/0499: Feedforward networks - G06N 3/08: Learning methods - G06N 3/09: Supervised learning - G16C 20/30: Prediction of properties of chemical compounds, compositions or mixtures - G16C 20/70: Machine learning, data mining or chemometrics - G16C: COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Disclosed herein is a polymer prediction system, comprising a deep learning neural network and a training dataset. The deep learning neural network can comprise: a property branch comprising two or more layers, each layer having a plurality of neurons; a polymer branch comprising two or more layers, each layer having a plurality neurons; and a merged layer including a concatenation operation, the concatenation operation configured to concatenate the property branch and the polymer branch. The training dataset can include a plurality of known polymers and a plurality of descriptors for each of the plurality of known polymers. Also disclosed herein are methods of using the same. ## Claims ### Claim 1 A method for predicting polymer properties comprising: analyzing a first polymer using a deep learning neural network to analyze a first set of functional groups of the first polymer, the deep learning neural network comprising a merged layer formed from the concatenation of a property branch and a polymer branch; determining a value from the merged layer; and outputting the value. ### Claim 2 (depends on 1) The method of claim 1 further comprising submitting the first polymer to a polymer prediction system; wherein the polymer prediction system comprises the deep learning neural network; wherein the determined value is indicative of a property of the first polymer; wherein outputting the value comprises outputting the value to a user of the polymer prediction system; wherein the property branch is a solubility branch; and wherein the property is a solubility of the first polymer in a first solvent. ### Claim 3 (depends on 2) The method of claim 2, wherein the value is indicative of the first solvent being a good solvent. ### Claim 4 (depends on 2) The method of claim 2, wherein the value is indicative of the first solvent being a nonsolvent. ### Claim 5 (depends on 2) The method of claim 2, wherein the determining further comprises determining; if the value is above a first predetermined threshold; and if the value is below a second predetermined threshold. ### Claim 6 (depends on 1) The method of claim 1 further comprising training the deep learning neural network using a training dataset comprising a plurality of known polymers and a plurality of descriptors for each of the plurality of known polymers. ### Claim 7 (depends on 1) The method of claim 1, wherein the property branch comprises two or more layers, each layer having from 1 to 100 neurons. ### Claim 8 (depends on 1) The method of claim 1, wherein the polymer branch comprises two or more layers, each layer having from 1 to 100 neurons. ### Claim 9 (depends on 1) The method of claim 1, wherein: one or both: the property branch comprises two or more layers; and the polymer branch comprises two or more layers; each layer has from 1 to 100 neurons; and each layer is constructed with a parameterized rectified linear unit. ### Claim 10 (depends on 1) The method of claim 1, wherein the merged layer comprises a sigmoid activation function. ### Claim 11 A deep learning neural network comprising: a property branch comprising two or more layers, each layer having a plurality of neurons; a polymer branch comprising two or more layers, each layer having a plurality neurons; and a merged layer including a concatenation operation, the concatenation operation configured to concatenate the property branch and the polymer branch. ### Claim 12 (depends on 11) A polymer prediction system comprising: the deep learning neural network of claim 11; wherein the deep learning neural network further comprises a training dataset comprising a plurality of known polymers and a plurality of descriptors for each of the plurality of known polymers; and wherein the system is configured to: receive an inputted first polymer having a first set of functional groups; analyze the first polymer using the deep learning neural network; determine a value for the first polymer, the value being indicative of a property of the first polymer; and output the property and the value to a user of the polymer prediction system. ### Claim 13 (depends on 12) The system of claim 12, wherein the property branch is a solubility branch; and wherein the property is a solubility of the first polymer in a first solvent. ### Claim 14 (depends on 12) The system of claim 12, wherein the value is indicative of the first solvent being a good solvent. ### Claim 15 (depends on 12) The system of claim 12, wherein the value is indicative of the first solvent being a nonsolvent. ### Claim 16 (depends on 12) The system of claim 12, wherein the system is further configured to determine; if the value is above a first predetermined threshold; and if the value is below a second predetermined threshold. ### Claim 17 (depends on 11) The deep learning neural network of claim 11, wherein each of the two or more layers in the property branch are constructed with a parameterized rectified linear unit. ### Claim 18 (depends on 11) The deep learning neural network of claim 11, wherein each of the two or more layers in the polymer branch are constructed with a parameterized rectified linear unit. ### Claim 19 (depends on 11) The deep learning neural network of claim 11, wherein the deep learning neural network has a latent space dimensionality selected from 3 to 10. ### Claim 20 (depends on 11) The deep learning neural network of claim 11, wherein the merged layer further comprises a sigmoid activation function.