Final ppt v3
Download as PPTX, PDF1 like387 views
This document summarizes a project to develop a structure-property linkage model for glass fibre reinforced polymer (GFRP) composites. The project involves collecting micro-CT scan data from GFRP samples, segmenting the scans into fibre and matrix phases, simulating microstructures and properties using finite element analysis, extracting 2-point statistics from the microstructures, using principal component analysis for dimensionality reduction, and developing a regression model to link microstructure properties to predict physical properties. The initial model is validated against a limited number of samples and further work is suggested to improve the model with more real samples and experimental testing.
![12/8/2014 2
Background
• Combine properties of the two materials
Figure 1: Components of a BMW sedan fabricated with
lignocellulosic fiber reinforced polymer (FRP)
composites [1] Figure 2: FRP materials in passenger
aircraft [2]
• Composites are used in many industries
Figure 3: Indian glass fibre composites
market (2006) [3]](https://image.slidesharecdn.com/finalpptv3-141208094721-conversion-gate01/85/Final-ppt-v3-2-320.jpg)
![12/8/2014 4
Motivation
• The structure and organization of the
reinforcement greatly affect the final
properties of the part
• Conventional approaches to property
determination do not take into account
the microstructure of the reinforcement
• Voigt model [4]
• Ruess Model [5]
• Those that do are based on a assumed
configurations of the fibres, not the
actual microstructure [6]
Figure 5: Complex microstructure of FRPC.](https://image.slidesharecdn.com/finalpptv3-141208094721-conversion-gate01/85/Final-ppt-v3-4-320.jpg)
![12/8/2014 10
Project Execution
Step 5: 2-Point Statistics
• 2-Point Statistics: Probability density associated with finding local states h and
h’ at the tail and head, respectively, of a prescribed vector r randomly placed
into the microstructure[7]](https://image.slidesharecdn.com/finalpptv3-141208094721-conversion-gate01/85/Final-ppt-v3-10-320.jpg)
![12/8/2014 11
Project Execution
Step 6: Dimensionality Reduction
• 2-Point Statistics: Extremely large set
• Low dimensional representation
• Principal Components Analysis [7]
• Linear transformation of high-dimensional data to a new orthogonal frame](https://image.slidesharecdn.com/finalpptv3-141208094721-conversion-gate01/85/Final-ppt-v3-11-320.jpg)
Final ppt v3
- 1. ME 8883/CSE 8803: Materials Informatics 8th December 2014 Developing Structure-Property Linkage for Glass Fibre Reinforced Polymer Composites Presented by Geet Lahoti, 2nd Year ISyE PhD Student Alicia White, 2nd Year MSE PhD Student Guided by Prof. Surya Kalidindi Dr. Tony Fast
- 2. 12/8/2014 2 Background • Combine properties of the two materials Figure 1: Components of a BMW sedan fabricated with lignocellulosic fiber reinforced polymer (FRP) composites [1] Figure 2: FRP materials in passenger aircraft [2] • Composites are used in many industries Figure 3: Indian glass fibre composites market (2006) [3]
- 3. 12/8/2014 3 Background • Forming processes create varied and complex microstructures • Microstructure varies even within a simple part such as this plate • Understanding the complexity of these microstructures is an open field which can give insight to the properties of these materials Figure 4: Variety in microstructure across an injected part
- 4. 12/8/2014 4 Motivation • The structure and organization of the reinforcement greatly affect the final properties of the part • Conventional approaches to property determination do not take into account the microstructure of the reinforcement • Voigt model [4] • Ruess Model [5] • Those that do are based on a assumed configurations of the fibres, not the actual microstructure [6] Figure 5: Complex microstructure of FRPC.
- 5. 12/8/2014 5 Project Outline • Objective: Develop Structure-Property Linkage for GFRPs Manufacture GFRP Samples Segmentation Is the no. of samples enough? Spatial Correlation Microstructure Simulation Dimensionality Reduction Perform Micro-computed tomography (micro-CT) Physical Property from Finite Element Analysis Physical Property from Experimental Testing Relationship Modelling Yes No
- 6. 12/8/2014 6 Project Execution Step 1: Samples and Micro-CT Data • Fibre: Glass • Polymer: Polypropylene • Processing: hot melt impregnation and extrusion/compression molding • Micro-CT Images: DICOM Format • No. of Samples: 2 • Dimensions of each sample: 1300 X 1300 X 900 voxels • Dimensions under consideration: 300 X 300 X 300 voxels
- 7. 12/8/2014 7 Project Execution Step 2: Segmentation • Need to separate the fiber from the matrix to get an accurate representation of microstructure • Apply peak fitting algorithm to histogram of pixel values • Segmentation based on Gaussian Likelihood Maximization • Gaussian Function 1 푎 f(x) = (푥−푏)2 2푐2 푒− where, a=height, b=center, c=width • Multi Otsu’s Method Original Microstructure Segmented Microstructure
- 8. 12/8/2014 8 Project Execution Step 3: Microstructure Simulation Fibers Elongated along Y Axis Fibers Elongated along Diagonal Fibers Elongated along X Axis Dimensions: 21 X 21 X 21
- 9. 12/8/2014 9 Project Execution Step 4: Physical Property Simulation • Finite Element was performed under uniaxial strain conditions • Property under consideration is going to be C11. • Stress and strain were calculated and used to find the components of the stiffness tensor corresponding to Ɛ1 Stress Strain
- 10. 12/8/2014 10 Project Execution Step 5: 2-Point Statistics • 2-Point Statistics: Probability density associated with finding local states h and h’ at the tail and head, respectively, of a prescribed vector r randomly placed into the microstructure[7]
- 11. 12/8/2014 11 Project Execution Step 6: Dimensionality Reduction • 2-Point Statistics: Extremely large set • Low dimensional representation • Principal Components Analysis [7] • Linear transformation of high-dimensional data to a new orthogonal frame
- 12. 12/8/2014 12 Project Execution Step 6: Dimensionality Reduction • Please add description about 2-Point Stats PCs for Sample 1: 0.0104 0.0000 0.0000 -0.0003 -0.0001 PCs for Sample 2: 0.0104 0.0000 0.0001 0.0002 0.0002
- 13. 12/8/2014 13 Project Execution Step 7: Structure-Property Linkage • Regression Property = ퟑퟑ. ퟔ +0.88 PC1 + 15.76 PC2 + 3.83 PC3 - 5.83 PC4 + 17.22 PC5 Rsquare: 0.9638 CV Mean Absolute Error: 0.14237 Property Predicted by model for sample 1 : 33.6129 and for sample 2: 33.6098 Property Predicted by FEM simulations for sample 1: 4.99 and for sample 2: 6.48
- 14. 12/8/2014 14 Conclusion & Future Work • Investigated digital representations of sample microstructures • Developed a S-P linkage based on simulated dataset • Obtain more real samples • Validating linkages with the segmented real microstructures • Carry out physical experimental testing of samples • Simulate a rich set of microstructures • Other Studies using the same protocol: Consider other composites like Carbon Fibre Reinforced Polymers
- 15. 12/8/2014 15 References 1. http://www.scielo.br/scielo.php?script=sci_arttext&pid=S1517-70762010000200006 2. http://www.reinforcedplastics.com/view/4437/india-on-the-up/ 3. http://essaywritingserviceuk.co.uk/advice-and-guidance/free-essays/the-potential-of-frp-materials- in-a-passenger-aircraft-structure/ 4. W. Voigt, ”On the relation between the elasticity constants of isotropic bodies," Ann Phys Chem 274 (1889): 573-587. 5. A. Reuss and Z. Angrew, ”A calculation of bulk modulus of polycrystaliine materials." ZAMM- Journal of Apllied Mathmatics and Mechanics, Vol. 9, No. 1, 1929, pp.49-58 6. http://onlinelibrary.wiley.com/doi/10.1002/pc.20002/pdf 7. Surya R. Kalidindi, “Data Science and Cyberinfrastructure: Critical Enablers for Accelerated Development of Hierarchical Materials
- 16. 12/8/2014 16 Thank You! Questions?
Editor's Notes
- #3: Composites are 2 component materials that are used to combine the properties of the two materials. Our composites are made of polypropylene reinforced with glass fibers. Combining these materials increases the strength of the material while maintaining light weight and mold ability. As a result, these materials are used in many industries to create light weight parts. The ability to predict the properties of the useful materials allows rapid development of such parts.
- #4: A complication to being able to predict the properties of composites is the rich variety of microstructures. Even within a simple part, there is variation of the microstructure caused by the forming methods. This image show in a location near the injection location, the fibers are highly disordered, whereas farther away the fibers have aligned to a greater degree.
- #5: Conventional approaches for? What do you mean by theoretical representation? Could you please add at least one theoretical model with reference? Add figure description The structure and organization of the reinforcement greatly affect the final properties of the part. As an example, the response of a part with aligned fibers will be very different if strain is applied in the fiber direction and perpendicular to it. Many often used approaches for determining the bulk properties are based solely on volume fraction. For example, the voigt and Ruess models are used to determine bounds on what the mechanical properties can be. However, to get more accurate predictions, more knowledge of the actual structure is necessary. Other researchers have attempted to look at similar systems by using distributions of characteristics. For example one study on natural fibers as reinforcement used probability density function for the fiber lengths and diameters. T
- #6: * Let me know if something doesn’t make sense. Real Microstructures Segmentation/Digitization Simulated Microstructures and Property Data Spatial Correlation (2-point Statistics) Dimensionality Reduction (PCA) Regression Model Nice chart
- #7: DICOM: Digital Imaging and Communications in Medicine
- #8: Simple thresholding doesn’t work since the modes of histogram are not clearly distinguishable! Otsu: maximizing the between-class variance! Gaussian Likelihood Maximization is something we made up explain it.
- #9: What variety do we have within each sample type? Fiber length, volume fraction? Within what range?
- #10: Please correct this slide. Add suitable values from the property you calculated. This figures are corresponding to Elongated Fibers (along X). If they doesn’t make sense let me know asap. In order to simulate the properties of the composite, finite element analysis was performed under uniaxial strain conditions. From the results, the first components of the stiffness tensor were calculated. For each sample C11 was used as a material property for use in creating the linkages.
- #11: Two point statistics is a method of capturing the microstructure information in a statistical way. This is an auto correlation so measuring the likelihood that fiber is at both ends of a vector of a certain length. Directionality can be seen in the two elongated samples. The diagonal sample requires partitioning the 3D data in at an angle for visualization. The maximum values also show the volume fractions.
- #15: In this project we have taken two microstructures generated from micro-CT data. We have investigated ways of using those microstructures to predict properties of the composite material. To validate this process we have used generated microstructures to create linkages between the structure of the sample and the properties. Future work on this project will include: continued investigation into the physical samples and use the results of that investigation to improve and validate the linkages we created. Other materials systems can then be investigated using this same protocol.
- #16: Why ref 4 and 5 are in German?? Reference no. 6 doesn’t work. Check the link. Fixed