1 / 38

Materials Informatics: Accelerating the Pace of Nano-Manufacturing Process Development

Materials Informatics: Accelerating the Pace of Nano-Manufacturing Process Development Curt M. Breneman*, Mike Krein, Rick Barto and Jason Poleski Nanomaterials for Defense Conference, May 6, 2010. What is Materials Informatics?. According to Wikipedia….

rumor
Download Presentation

Materials Informatics: Accelerating the Pace of Nano-Manufacturing Process Development

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Materials Informatics: Accelerating the Pace of Nano-Manufacturing Process Development Curt M. Breneman*, Mike Krein, Rick Barto and Jason Poleski Nanomaterials for Defense Conference, May 6, 2010

  2. What is Materials Informatics? According to Wikipedia… • Involves the collection, storage and use of materials information • Holds the promise of prospective materials property prediction… Brinson Advanced Materials Lab

  3. Materials Informatics:Significance and Context • Unified vision of Cheminformatics and Materials Informatics • In terms of fundamental physics of molecular interactions • Builds upon expertise and capabilities of RECCR • Leverages corporate datasets, experimental efforts

  4. Structure of Today’s Talk Traditional Approaches Materials Informatics The Future

  5. Traditional Approaches • “Cut-and-try” strategies were used to find better materials for specific applications • Progress depended on Local Expertise • “Off the shelf” materials improved slowly over time

  6. Design of Application-Specific High-performance Materials Materials with exotic properties are needed Multiple-property optimization is required Existing design methods fail to address these issues This is the driving force for the Paradigm Shift

  7. The Paradigm Shift… • There is a rich history of Cheminformatics development in drug discovery • Enabling advances have taken place in computer software, hardware and data collection methods • Hard lessons were learned - particularly about Descriptors, Model Validation and Domains of Applicability • Then a Key Observation: • Given appropriate descriptors, predictive models of materials properties should be available through the application of these same advances… • The Tipping Point: Materials Informatics can streamline design - • Realization that Cheminformatics approaches to materials design will become standard engineering tools • Cheminformatics modeling and validation methods can be used to optimize materials with exotic properties

  8. Materials Quantitative Structure-Property Relationships (MQSPR) Successful (M)QSPR methods can: Find "hidden" information in chemical data Retrieve structures with well-defined property ranges from a database Identify appropriate descriptor types for each situation Assess model applicability domains Validate modeling results

  9. Quantitative Structure-Property Relationship (QSPR) Modeling • Analysis of variance and identification of relationships within data

  10. Structure-Property Relationships X PROPERTY STRUCTURE Statistical or Pattern Recognition Methods Computational Chemistry MOLECULAR DESCRIPTOR REPRESENTATION

  11. Structure-Property Relationships X PROPERTY STRUCTURE Data Acquisition Statistical or Pattern Recognition Methods Encoded Process Design Parameters

  12. Capture experimental data in one central repository Provides centralized web-based tool for data review and analysis Build repository with considerable input from experimentalists GUI-based analysis interfaces included Perform analysis on experimental data to provide additional guidance Visualization and clustering, linear and nonlinear modeling Next Steps: Build-out analysis capabilities, robustness, GUI design Provide inputs for optimal experimental design Technical Approach

  13. Modeling and Validation DATASET Training set Test set Real Y-scrambling model validation! Bootstrap sample k Predictive Model Training Validation Scrambled Learning Model Tuning / Prediction Prediction

  14. Simple models are better Interpretable models are better Reality: need to balance predictive ability and interpretability Model Parsimony Rules:

  15. Linear and Nonlinear Methods Data Preprocessing Model / Method Validation Domain of Applicability Assessment “Automated” Modeling

  16. http://reccr.chem.rpi.edu/ONR/ “Automated” Modeling

  17. System 1 Modeling: PLS

  18. System 1 Modeling: SVM

  19. System 1 Modeling: RF

  20. System 1 Sensitivity AnalysisPLS

  21. System 1 Sensitivity AnalysisSVM

  22. System 1 Sensitivity AnalysisRandom Forest

  23. System 1 Modeling:Feature Sensitivity Plots

  24. System 1 Modeling:Feature Sensitivity Plots

  25. System 2 Modeling: PLS

  26. System 2 Modeling: SVM

  27. System 2 Modeling: RF

  28. System 2 Sensitivity AnalysisPLS

  29. System 2 Sensitivity AnalysisSVM

  30. System 2 Sensitivity AnalysisRandom Forest

  31. System 2 Modeling:Feature Sensitivity Plots

  32. System 2 Modeling:Feature Sensitivity Plots

  33. Modeling Workflow Data Analysis Results Experiment Results Multivariate Analyses Inspect Model Results Predict Unknown Cases Experiment Selection Scrutinize Validity of Model Validation /Domain Assessment Test Predictions Data Reduction Expand Database & “Learn” Evaluate, Reselect Data or Descriptors Data Clustering Human Task Descriptor/ Feature Selection Modeling Software Task Legend Satisfactory Unsatisfactory

  34. Smart Material Manufacturing • What to make next? • How to make it? • With what confidence? Results Predict Unknown Cases Test Predictions Expand Database & “Learn”

  35. Inverse QSAR Approach • First Steps: Beyond “Sensitivity Analysis”: (Now) • Build models from existing data and find important parameters • Alter identified parameters for the best experimental data points • Make predictions on “new” data points using existing models • Create materials and compare properties with predictions Validation Do Prediction Run Experiment If Results Don’t Match Prediction, Why?

  36. Inverse QSAR Approach • Next Steps: Making it Useful ! (Just on horizon) • • The most interesting materials usually •    aren’t the best performing in one category • • User selects the experimental parameters to alter for “what if” analysis • • Logical extension of this is multi-objective QSAR and iQSAR Validation Do Prediction Run Experiment If Results Don’t Match Prediction, Why?

  37. The RECCR Community http://reccr.chem.rpi.edu

  38. ACKNOWLEDGMENTS Current and Former members of the RECCR/DDASSL group Breneman Research Group (RPI Chemistry) N. Sukumar M. Sundling Min Li Long Han Jed Zaretski Taowei Huang Theresa Hepburn Mike Krein Steve Mulick Shiina Akasaka Hongmei Zhang C. Whitehead (Pfizer Global Research) L. Shen (BNPI) L. Lockwood (Syracuse Research Corporation) M. Song (Synta Pharmaceuticals) D. Zhuang (Simulations Plus) W. Katt (Yale University chemistry graduate program) Q. Luo (J & J) Embrechts Research Group (RPI DSES) Tropsha Research Group (UNC Chapel Hill) Bennett Research Group (RPI Mathematics) Collaborators: Brinson Advanced Materials Lab – (Northwestern) Schadler Research Group (RPI Materials Engineering) Funding Office of Naval Research (ONR) Lockheed-Martin (LMCO) NIH (GM047372-07) NIH (1P20HG003899-01) NSF (BES-0214183, BES-0079436, IIS-9979860) GE Corporate R&D Center Ford-Boeing Alliance NSF NIRT Program Chemical Computing Group (CCG)

More Related