Systems and methods for formulating material in a data-driven manner

Abstract

Systems and methods for optimizing the formulation of materials are provided. The systems and methods employ a data-driven, iterative approach to derivate optimal material formulations. One portion of the system includes a sample automation system that outputs the material samples to be tested, and a second portion of the system includes an optimization engine that analyzes data extracted from the material samples and generates additional formulations for materials to be printed and tested. This process continues so that optimal material formulations can be determined based on desired mechanical properties of the material to be optimized. The optimization engine can further be capable of predicting results of formulation that have not yet been tested and using those predictions to further drive the next suggested materials to be tested.

1 . A method of formulating a material, comprising:
 dispensing a plurality of material samples, each of the plurality of material samples having different properties;   measuring data related to the plurality of material samples;   determining a suggested plurality of material samples based on the measured data related to the plurality of material samples, the suggested plurality of material samples comprising material samples having a plurality of different formulations selected from a design space comprising a set of possible formulations, the determining comprising:
 generating, for each of a plurality of performance objectives, predicted performance characteristics of the design space based on the measured data, 
 determining, for each of the plurality of performance objectives, a predicted Pareto front based on the predicted performance characteristics of the design space, and 
 selecting the plurality of different formulations from untested regions of the design space based on the predicted Pareto front for at least one of the plurality of performance objectives; and 
   dispensing a subsequent plurality of material samples based on the suggested plurality of material samples.   
     
     
         2 . The method of  claim 1 , wherein determining a suggested plurality of material samples based on the measured data related to the plurality of material samples, further comprises:
 optimizing the plurality of different formulations of the suggested of material samples based on attempting to maximize each of the plurality of performance objectives simultaneously, resulting in a determination of an overall Pareto front for a material to be dispensed.   
     
     
         3 . The method of  claim 2 , wherein optimizing a formulation of the suggested plurality of material samples based on attempting to maximize each of the plurality of performance objectives simultaneously, further comprises utilizing a Bayesian optimization strategy to optimize the plurality of different formulations of the suggested material samples based the plurality of performance objectives. 
     
     
         4 . The method of  claim 3 , wherein utilizing a Bayesian optimization strategy to optimize the plurality of different formulations of the plurality of suggested material samples based on the material toughness, compression modulus, and maximum strength of the plurality of suggested material samples, further comprises:
 fitting a Gaussian Process for each performance objective of the plurality of performance objectives independently;   performing Thompson sampling of the Gaussian Process for each of plurality of performance objectives;   approximating a predicted Pareto set and the predicted Pareto front for each of the plurality of performance objectives; and   outputting the plurality of different formulation of the suggested plurality of material samples based on the predicted Pareto sets and the predicted Pareto fronts.   
     
     
         5 . The method of  claim 4 , wherein determining a suggested plurality of material samples based on the measured data related to the plurality of material samples, further comprises:
 predicting one or more of material toughness, compression modulus, and maximum strength for one or more material samples not yet dispensed; and   factoring in such one or more samples not yet dispensed in outputting the plurality of different formulation of the suggested plurality of material samples.   
     
     
         6 . The method of  claim 5 , wherein predicting one or more of material toughness, compression modulus, and maximum strength for one or more material samples not yet dispensed further comprises:
 using a Bayesian optimization strategy.   
     
     
         7 . The method of  claim 1 , wherein measuring data related to the plurality of material samples further comprises:
 performing compression testing.   
     
     
         8 . The method of  claim 7 , further comprising:
 using a stress-strain curve generated by the performing compression testing to generate at least one of material toughness, compression modulus, or maximum strength of the plurality of material samples.   
     
     
         9 . The method of  claim 1 , further comprising:
 mixing the dispensed plurality of material samples prior to measuring data related to the plurality of material samples.   
     
     
         10 . The method of  claim 9 ,
 wherein the plurality of material samples comprise a plurality of formulation primaries, and   wherein mixing the dispensed plurality of material samples prior to measuring data related to the plurality of material samples is based on a selected ratio for the plurality of formulation primaries to create at least one material sample of the plurality of material samples.   
     
     
         11 . The method of  claim 10 , wherein the selected ratio is based on the suggested plurality of material samples. 
     
     
         12 . The method of  claim 11 , wherein the plurality of formulation primaries are configured for use as 3D printing ink. 
     
     
         13 . The method of  claim 9 , wherein after mixing the dispensed plurality of material samples but prior to measuring data related to the plurality of material samples, the method further comprises:
 fabricating the mixed plurality of material samples; and   performing one or more post-processing procedures on the mixed plurality of material samples to normalize each material sample of the mixed plurality of material samples, the mixed plurality of material samples being the plurality of material samples from which the data is measured.   
     
     
         14 . The method of  claim 13 , wherein at least one of: dispensing a plurality of material samples, mixing the dispensed plurality of material samples prior to measuring data related to the plurality of material samples, fabricating the mixed plurality of material samples, or performing one or more post-processing procedures on the mixed plurality of material samples is performed in a semi-automated manner. 
     
     
         15 . The method of  claim 13 , wherein at least one of: dispensing a plurality of material samples, mixing the dispensed plurality of material samples prior to measuring data related to the plurality of material samples, fabricating the mixed plurality of material samples, or performing one or more post-processing procedures on the mixed plurality of material samples is performed in a fully automated manner. 
     
     
         16 . The method of any of  claim 1 , wherein dispensing a plurality of material samples comprises dispensing a plurality of material samples in which the samples have different properties. 
     
     
         17 . The method of any of  claim 1 , wherein determining a suggested plurality of material samples based on the measured data related to the plurality of material samples comprises determining a suggested plurality of material samples in which the suggested samples have different properties. 
     
     
         18 . The method of  claim 1 , wherein determining further comprises determining a quality of a Pareto front for each of the plurality of performance objectives. 
     
     
         19 . The method of  claim 18 , wherein selecting the plurality of different formulations further comprises selecting the plurality of different formulations based on at least one of: a predicted improvement of the quality of the Pareto front for at least one of the plurality of performance objectives or a predicted reduction in an uncertainty of at least one of the predicted performance characteristics of the design space. 
     
     
         20 . The method of  claim 2 , wherein the plurality of performance objectives comprises material toughness, compression modulus, and maximum strength.