In-situ Electronics Measurement Using X-ray Micro-Computed Tomography System and Data Driven Prognostic Health Management
View/ Open
Date
2017-05-02Type of Degree
PhD DissertationDepartment
Mechanical Engineering
Metadata
Show full item recordAbstract
In the 21st century, the electronics is everywhere and covers every aspect of human life. You can find it when watching TV, opening a laptop, or checking emails on a cell phone. Since 2015, 268 million PCs, 455 million tablets, and 253.1 million TVs have been sold globally. Despite of the power of electronics, the failures of electronics would cause a variety of issues. One of the issues is electronic waste. When an electronics breaks down, it will be more likely to be replaced than repaired, because the cost of repair is usually much higher than buying a new device. There are two approaches to resolve this issue. On the one hand, we increase the reliability of electronics to maximize their lives. On the other hand, we study the Remaining Useful Life (RUL) of electronics, which provides life expectancy information of electronics. Knowing the RUL increases the efficiency of electronics usage in that it helps people avoid unnecessary labor of repairing when electronics approaches its designed life. The In-situ deformation of electronics affects their reliability, which could be used as the failure-leading indicator. The current dissertation focuses on the reliability of electronics, and how to quantitatively assess it during the operation. The reliability of electronics can be affected by many factors including operational environment, manufacture, and frequency of usage. All of these factors contribute to mechanical failure of electronics, which can be reckoned as one of the major failure of electronics.. The mechanical failures of electronics can result in (a) encapsulate failure, (b) connector failure, (c) open circuit failures and (d) short circuit failures, and (e) component failure. The primary cause of mechanical failure is deformation. Therefore, it is very import to know how much deformation is generated in the electronics during its operation. Previously, the topic of deformation has been studied by many scholars. Several techniques are available to investigate the deformation of an object including strain gages, moiré method, and digital image correlation. The strain gages and moiré method require a pre-surface preparation, which creates a surface attachment or surface coating. The Digital Image Correlation(DIC) is an advanced technique that has been developed recently, and it has already been used widely to measure the surface deformation and strain. The current research focuses on the three-dimensional (3D) deformation field, which provides a full field deformation and strain. Specifically, we apply the Digital Volume Correlation (DVC) technique to investigate the In-situ deformation of electronics. The development of the DVC technique began with Bay, Smith, Fyhrie, and Saad, 1999. The article illustrated the basic concepts and applications on the measurement of 3D displacements and strain in the Trabecular Bone using X-rays 3D reconstructions, under 35 microns voxel sensitivity . Since then, there have been many DVC applications in various fields, using a variety of data acquisition system and under different voxel sensitivities (Lenoir & Bornert, 2007) (Roux, Hild , Viot, & Bernard, 2008) (Haldar, Gheewala, & Grande-Allen, 2011) (Hussein & Barbone, 2012) (Moegeneyer & Helfen, 2013) (Brault, Germaneau, & Dupre, 2013) (Germaneau & Doumalin) (Haase & Noack, 2014) (Leclerc, Roux, & Hild, Projection Savings in CT-based Digital Volume Correlation, 2014). To begin with, the digital volume data can be acquired by the Scatter Light (Germaneau, Doumalin, & Dupre, 2007), and in this application the voxel sensitivity is about 60 micrometers. More recently, the application of micro-MRI (Benoit & Guerard, 2009) processes a larger sized 3D volume of 512-256-256 voxels and 40-20-20 mm in each dimension, for which the voxel sensitivity is about 78 micrometers. Latest the laser scanning confocal microscopy (LSCM) (Toyjanova, et al., 2014) uses volume with size 512-512-P voxels with P range from the 64 to 256 voxels (i.e. 20 to 77 micrometers). The LSCM is also of the highest voxel sensitivity under the optical light condition, which could process organic tissues and living cells. In summary, all of these methods could generate gray-scale based digital volumes whose sizes vary according to the region of interest (ROI). Despite of the development of DVC techniques, one of the issues regarding DVC application is its computational cost. The larger the size of the digital volume, the higher the computational cost will be. The paper of Gates, Lambros, and Heath (2012) proposed to increase the computational efficiency of DVC application using Cross-Correlation with the FFT algorithm. The paper implements the parallel computation technique to speed up the processed time. The current study incorporated Gates et al. (2012)’s method in to calculate the deformation of electronics from the X-ray CT scans using the DVC technique, which enables us to compute the full field strains in the package efficiently. In the meanwhile, we also create a 3D data visualization tool to view the results in differential 3D layers. In Chapter 3, we showed the general 3D image-processing framework. Some applications such as defect and failure detection (Lee & Nguyena, 2000) were established using this visualization technique. Then, we introduced the DVC algorithm and its development. A detailed description of DVC algorithm would be presented in the Chapter 4. In Chapter 5, several applications using the DVC were presented. First of all, we demonstrated the proof-of-concept with three points bending test using aluminum beam. Secondly, we applied the theorem along with the DVC technique to measure the In-situ LED chip deformation caused by the CTE mismatch. Then, we showed that the deformation could be measured in the operational BGA package. In the latest development, we introduced this technique to measure the soft material deformation, in which we solved several issues that adapting the DVC algorithm to the soft-matter and low-density materials. In Chapter 6, we introduced an advance Finite Element Method(FEM), which is based on the mesh generated from the X-ray CT scan. Using the 3D volumetric image, the geometry of object can be reconstructed from the ISO-surfaces, and we can create the FEM discretization within those surfaces. The FEM discretization utilizes the four vertices and four faces tetrahedron elements. Then, this mesh work can be used with the FEM simulation. Moreover, in chapter 6, we demonstrate a thermo-mechanical simulation on the BGA package, in which the defects such as solder void and crack has been presented. In the end, the cumulative plastic strain energy has been computed, and the stress-strain hysteresis loop has also been computed. In Chapter 7, we used another technique to track and to predict the life of electronics--the Data Driven Prognostication technique, which finds and tracks leading failure indicators of a system. In electronics systems, the electronic resistance is usually picked as leading indicator for the mechanical failures because by Ohm’s Law, resistance will increase with the presence of a solder crack. Similarly, in LED system we use the lumen maintenance as a leading indicator for mechanical failures, because the lumen maintenance output will decrease as the system degrades. Using the data driven methods (i.e., Kalman Filter, Extended Kalman Filter, and Bayesian Method), we iteratively update the system state vectors to find the best predictor of RUL. Then, the RUL can be predicted and calculated from the extrapolations using characteristic function. Overall, in these two systems, the PHM framework has been established, and we will cover this in details in the Chapter 7.