A computationally efficient approach for inverse material characterization combining Gappy POD with direct inversion
Introduction
Inverse problems relating to the characterization of various material properties of a variety of solids/structures and systems are of paramount importance in a wide range of science and engineering fields. For instance, structures, from industrial to biological, could be evaluated to determine their current state of health based upon their material properties, whether mechanical, thermal, or electrical. Corresponding to this substantial interest in material property characterization, a wide variety of inverse problem solution strategies have been developed relating to a variety of applications, such as structural health monitoring and nondestructive evaluation [1], [2] and biomechanical imaging [3], [4], [5], [6], [7], among other applications [8], [9].
Since it is often not possible to find analytical solutions for inverse problems in practice, due to problem ill-posedness and/or complexities in geometry and boundary conditions, among other challenges, computational inverse characterization solution approximation approaches have become common. Overall, computational inverse characterization approaches, which are typically based around some type of computational representation of the mechanics of the system of interest (e.g., finite element analysis), have been shown through several studies [9], [5], [3] to provide generalized frameworks for treating and distinguishing between various contributions to a system response, while providing physically meaningful solutions that can be applied to predict future behaviors. However, there is as wide a variety of computational inverse problem approaches as there are applications, with each having different strengths and weaknesses, and their effectiveness is significantly dependent upon the specifics of the application of interest. Moreover, the different computational approaches often differ significantly in the tradeoff between computational efficiency and solution accuracy.
One general way that computational inverse problem solution approaches can be divided is into those that use iterative optimization and those that are non-iterative/direct. Typically, the optimization-based approaches attempt to determine the unknown properties that minimize the difference between the measured system response and the response predicted by the computational representation of the mechanics of the system [2], [10], [11]. The optimization approaches can be divided further into those that use gradient-based optimization (e.g., Newton’s method, conjugate gradient, etc.) and those that use non-gradient-based optimization (e.g., random search, genetic algorithm, etc.) to minimize the response error. Gradient-based optimization methods typically require substantially less iterations to converge to a solution approximation in comparison to non-gradient-based methods (i.e., are less computationally expensive), but often become trapped in local minima (i.e., an inaccurate solution), while non-gradient-based (since they commonly include some stochastic component) often have closer to global search capabilities. Alternatively, non-iterative methods attempt to somehow directly relate the measured response to the unknown parameters [12], [13]. As such, once set up, non-iterative methods are naturally almost negligible in computing cost in comparison to the optimization-based approaches.
Non-iterative methods include machine learning approaches that use experimentally and/or numerically generated datasets of potential inverse problems solution parameters and the corresponding system responses to train a “surrogate” mapping (e.g., artificial neural network) to approximate the relationship between the system response (as inputs to the mapping) and inverse unknowns (as outputs of the mapping) [14], [15]. Then, provided with a measured system response, the surrogate mapping can estimate inverse solution parameters in fractions of a second. Although relatively simple to implement, these machine learning approaches completely exclude any knowledge of the mechanics of the system beyond what can be naturally discerned from the dataset, and can have significant problems creating the surrogate mapping at all, since the input–output relationship is often not one-to-one, leading to substantial accuracy concerns in many cases. Alternatively, there are several different approaches (referred to as “direct inversion” methods herein) that instead manipulate the equations governing the mechanics of the system (i.e., the forward boundary value problem) to create a solution process similar to that of solving the forward problem itself (e.g., similar in process to a finite element analysis to predict the deformation response of a solid given geometry, material properties, and boundary conditions). Therefore, the direct inversion methods provide a solution estimate based upon the measured response and the mechanics assumed to govern the response of the system considered at a cost on the order of a single numerical analysis of the forward problem.
Direct inversion approaches have seen considerable application in problems relating to characterization of the distribution of elastic properties from mechanical testing. Furthermore, these approaches include those that solve for the material properties over the entire (or almost the entire) domain at once (i.e., global methods), as well as those that break the domain into subregions and determine the properties for each subdomain in sequence (i.e., local). Global methods typically use some variation of either the finite difference method (FDM) or the finite element method (FEM) to create a system of equations based on the forward boundary value problem to solve the node or element-based elastic modulus distribution everywhere in the domain. FDM approaches for global direct inversion include the seminal work of Raghavan and Yagle [16] for simultaneously characterizing stiffness and hydrostatic pressure distribution of tissue from strain measurements, as well as other subsequent similar approaches for characterization of relative elastic modulus [17] and relative shear modulus [18] distributions. FEM approaches have shown some benefits over FDM approaches for global direct inversion in that they can be cast in a way so as to only require the gradient of displacement (i.e., a single derivative) rather than the divergence of strain (i.e., two derivatives of displacement). In this way, FEM approaches can be more tolerant to noise and other measurement errors than FDM approaches. Examples of FEM approaches for global direct inversion include the work by Zhu et al. [12] for elasticity reconstruction from displacement measurements and the work by Park and Maniatty [13] for shear modulus reconstruction from measured steady-state dynamic displacement fields. The local methods are often similar to the global methods, with several approaches similarly using principles from FDM and FEM (although not all), with the fundamental difference being just how the system is discretized/solved. Oliphant et al. [19], [20] presented a local approach that used polynomial fit of the measured dynamic displacement field and the strong form of the governing equation of motion to characterize the elasticity distribution in tissues, while Romano et al. [21], [22] used a weak form FEM-type approach for element-by-element characterization of the ratio between both Lamé constants and density from dynamic displacement measurements. In a further modification, Albocher et al. have proposed an adjoint-weighted equation approach based on Galerkin discretization, which results in a stable and convergent direct inversion method for estimating material property distributions [23]. One particularly unifying aspect of the direct inversion approaches is that the entire (or nearly entire) spatial distribution (i.e., full-field) of the system response (e.g., displacement) must be measured/available to successfully characterize the unknown properties. Due to this requirement, the use of direct inversion (particularly of mechanical properties) has been mainly limited to biomechanical imaging applications (e.g., characterization of elastic properties of tissues from medical imaging data), where full-field or nearly full-field deformation information is regularly available. An additional common challenge of direct inversion approaches is their noise sensitivity. Most direct inversion approaches are not capable of producing a usable solution estimate with any significant level of measurement noise, which has resulted in several investigations into strategies for signal denoising prior to applying direct inversion [13], [20].
This work presents an approach to utilize measurement data from only a portion of the system domain (i.e., partial-field data) for direct inversion of material properties. In particular, the approach is presented in the context of characterizing the spatial distribution of the elastic modulus of solids (i.e., elastography) provided with displacement response measurements over some portion of the solid domain. The core component of this approach is the use of the Gappy proper orthogonal decomposition (POD) machine learning strategy to build a data reconstruction tool that can predict the full-field response of a system from the available partial-field measurements. This data reconstruction tool is built from a set of potential solution fields that are generated based upon information known a priori about the nature of the potential inverse problem solutions (e.g., arbitrarily generated approximate potential elastic modulus distributions). Once the full-field response reconstruction is complete, the full-field response is applied in a finite element-type direct inversion strategy to estimate the unknown material property distribution everywhere in the domain at a computational expense approximately equivalent to a single finite element analysis of the system. In addition, the Gappy POD approach can also act as somewhat of a noise filter during the reconstruction process, thereby, providing an added benefit of reducing the effects of measurement noise on the subsequent direct inversion solution procedure.
The following section presents the details of the Gappy POD approach to reconstruct full-field data from partial-field measurements. Then, Section 3 shows the direct inversion algorithm to calculate the spatial distribution of elastic modulus provided with the full-field displacement response and the boundary conditions corresponding to the test method used to produce the measurements (i.e., the constraints on the solid and the excitation used to generate the displacement measurements). The complete direct inversion with Gappy POD algorithm is summarized in Section 4. Section 5 presents two simulated case studies relating to characterization of localized elastic modulus distributions in solids to evaluate the capabilities of the inverse characterization procedure, which are followed by an example utilizing (actual) experimentally obtained displacement measurements to evaluate the stiffness distribution of an arterial construct to validate the “real-life” applicability of the approach.
Section snippets
Gappy proper orthogonal decomposition
Gappy POD is an extension of the traditional POD approach that was first developed and presented by Everson and Sirovich [24] for the purpose of filling in missing information to reconstruct marred photos. Subsequently, Gappy POD has shown substantial capabilities to accurately reconstruct physical processes from partial-field measurement data, especially for fluid flow problems [25], [26], [27], [28], [29], and has even been investigated for use in creating surrogate mappings (as described
Direct inversion of elastic modulus
As discussed, although potentially applicable to a variety of different physical systems, the application of the present work is characterization of the elastic modulus distribution of a solid from displacement measurements (full-field displacement response once Gappy POD has been utilized). Furthermore, the following formulation is presented with respect to a steady-state dynamic testing procedure (as could be applicable to frequency response function-based evaluation), but could easily be
Algorithm for direct inversion from partial-field measurements with Gappy POD
The overall algorithm for direct inversion of a material property distribution from partial-field response measurements with Gappy POD can be summarized as follows:
- Given:
The geometry of the structure of interest, the boundary conditions and partial-field response measurements from a nondestructive testing procedure, and any available material properties.
- Find:
The unknown material property distribution.
- Step 1:
Generate (e.g., randomly or through some other sampling procedure) a set of potential
Examples and discussion
Three examples were considered and evaluated to examine the potential benefits and capabilities of using the direct inversion approach with Gappy POD to characterize the elastic modulus distribution in solids from partial-field measurements. In all three cases, the partial-field measurements were assumed to be obtained from some type of nondestructive testing in the linear range of the solid behavior. First, two numerical examples were considered in which the “experiment” was simulated through
Conclusions
An approach for inverse material characterization that combines Gappy proper orthogonal decomposition with direct inversion for computationally efficient characterization with data measurements from only a portion of the system domain was presented and analyzed. The approach first uses the Gappy POD method to estimate the response field over the entire system domain from the available measurements along with any available a priori information regarding the potential solution distribution. Then,
Acknowledgment
The authors gratefully acknowledge the financial support of the Air Force Office of Scientific Research through Award No. FA9550-11-1-0132.
References (38)
- et al.
Inverse viscoelastic material characterization using pod reduced-order modeling in acoustic–structure interaction
Comput. Methods Appl. Mech. Engrg.
(2009) - et al.
Solution of the nonlinear elasticity imaging inverse problem: the incompressible case
Comput. Methods Appl. Mech. Engrg.
(2011) - et al.
Large scale parameter estimation problems in frequency-domain elastodynamics using an error in constitutive equation functional
Comput. Methods Appl. Mech. Engrg.
(2013) - et al.
Detection of cracks using neural networks and computational mechanics
Comput. Methods Appl. Mech. Engrg.
(2002) - et al.
Adjoint-weighted equation for inverse problems of incompressible plane-stress elasticity
Comput. Methods Appl. Mech. Engrg.
(2009) Unsteady flow sensing and estimation via the gappy proper orthogonal decomposition
Comput. & Fluids
(2006)- et al.
Non-invasive assessment of elastic modulus of arterial constructs during cell culture using ultrasound elasticity imaging
Ultrasound Med. Biol.
(2013) - et al.
Robust structural damage identification based on multi-objective optimization
Internat. J. Numer. Methods Engrg.
(2010) - et al.
Assessment of multi-objective optimization for nondestructive evaluation of damage in structural components
J. Intell. Mater. Syst. Struct.
(2014) - et al.
Inverse estimation of viscoelastic material properties for solids immersed in fluids using vibroacoustic techniques
J. Appl. Phys.
(2007)
Quantitative elasticity imaging: what can and cannot be inferred from strain images
Phys. Med. Biol.
Linear and nonlinear elasticity imaging of soft tissue in vivo: demonstration of feasibility
Phys. Med. Biol.
A source sensitivity approach for source localization in steady-state linear systems
Inverse Problems
Topological derivative for the inverse scattering of elastic waves
Quart. J. Mech. Appl. Math.
Solution of the nonlinear elasticity imaging inverse problem: the compressible case
Inverse Problems
A finite-element approach for Young’s modulus reconstruction
IEEE Trans. Med. Imaging
Shear modulus reconstruction in dynamic elastography: time harmonic case
Phys. Med. Biol.
Damage detection in a girder bridge by artificial neural network technique
Comput.-Aided Civ. Infrastruct. Eng.
Forward and inverse problems in elasticity imaging of soft tissues
IEEE Trans. Nucl. Sci.
Cited by (25)
Method to solve redundant inverse problems based on a Latent Semantic Analysis approach. Application to an aerojet engine
2020, Aerospace Science and TechnologyCitation Excerpt :With this approach, the user gets, typically, the answer to his/her query in less than a second. Proper orthogonal decomposition and gappy proper orthogonal decomposition for model reduction in the context of inverse problems have been used by Brigham and Aquino [1] and Wang et al. [2] for material characterization. In reference [1] the inverse problem was cast as an optimization problem and the solution was found by minimizing an error functional.
A novel inversion approach for identifying the shape of cavity by combining Gappy POD with direct inversion scheme
2020, International Journal of Heat and Mass TransferCitation Excerpt :These results show that the Gappy POD method has the advantages of reducing the order of the model and improving the computational efficiency. Based on these excellent properties, Gappy POD is used to inverse directly the material characterization [26]. Also, the research shows that Gappy POD can act as somewhat of a noise filter during the reconstructing full-field data.
Computational design optimization of a smart material shape changing building skin tile
2019, Engineering StructuresCitation Excerpt :The more general the parameterization the higher the computational expense of the problem. For example, finite element-type discretizations of a material property distribution [28], for which every node or element of a mesh can have a different property, have a high degree of general applicability. However, the large number of unknowns in a mesh description can substantially increase computational expense and may require some kind of regularization or other additional consideration to address ill-posedness.
Inversing heat flux boundary conditions based on precise integration FEM without iteration and estimation of thermal stress in FGMs
2019, International Journal of Thermal SciencesCitation Excerpt :More recently, Yu et al. [32,33] employed LSM and BEM to inverse the temperature and heat flux boundary conditions for steady and transient heat conduction problems. According to these literature, we find that the computational cost of non-iterative method is almost negligible in comparison with iterative method [34] and different types of boundary conditions can be obtained quickly and accurately by non-iterative inverse methods, which will lay the foundation for accurate estimation of thermal stress distributions. However, it is well known that the IHCP is ill-posed, and the inversed results may be highly sensitive to the inevitable measurement error [35] such as the random error on the measurement temperature and the bias on supposed known parameters [36,37].
On the value of test data for reducing uncertainty in material models: Computational framework and application to spherical indentation
2019, Computer Methods in Applied Mechanics and EngineeringA novel non-iterative method for estimating boundary conditions and geometry of furnace inner wall made of FGMs
2019, Applied Thermal EngineeringCitation Excerpt :The non-iterative method attempts to find the relationship between the information of measurement points and the unknown quantities to be inversed. As such, once set up, the non-iterative method is naturally almost negligible in computational cost in comparison to the iterative method [26]. In 1997, Yang [25] applied the linear LSEM to inverse the strength of the time-dependent heat source, in which the direct problem is solved by finite difference method (FDM).