A multi-objective memetic inverse solver reinforced by local optimization methods☆
Introduction
Parametric inverse problems (IPs) for partial differential equations (PDEs) consist of restoring the values of PDE parameters (inverse solutions) from the known observation of their solution called forward solution over certain subdomains. IPs are fundamental in several applications, such as in oil and gas exploration, structure health monitoring, and cancer tissue diagnosis (see e.g. [1]).
The most popular mathematical formulation of IPs is in terms of global optimization problems (GOPs), where the decision variables belong to the admissible set of parameter functions representations, and the objective functionals to be minimized express the misfit between measured and simulated PDE forward solutions.
Solving IPs meets many obstacles caused by their ill-conditioning. If the ill-conditioning involves the lack of the global misfit convexity only, but still there exists a unique global minimizer, the misfit regularization (see e.g. [2]) can be the effective way to obtain its numerical solution. If the problem possesses more than one solution (i.e. the misfit is multimodal and has many global minimizers) and/or the misfit is insensitive with respect to several decision variables in the neighborhood of the global minimizer, the complex stochastic searches (see e.g. [3]) allow to overcome the above difficulties when solving them.
The misfit multimodality and insensitivity are generally caused by the lack of information about the phenomenon to be analyzed. It may result in a mathematical formulation of the problem that allows multiple solutions (see e.g. [4]) or the uncertain misfit representation due to the irreducible measurement errors (see e.g. [5], [6], [7]). The other obstacles are caused by artifacts that might be produced by deterministic and stochastic global optimization strategies (see e.g. [8]).
The most straightforward way to improve the IPs conditioning is by increasing the amount of information about the studied phenomenon. If the phenomenon is composed of multiple physical processes, then it is possible to consider many misfits simultaneously, each one associated with a separate physics. This approach leads to the multiobjective global optimization problem defined and discussed in Section 1.2.
The idea of solving ill conditioned IPs by finding Pareto solutions for misfits imposed by multiple physics is rarely described in the literature. The authors of [9] apply the inverse quantitative structure–property relationship for designing new chemical compounds. Optimal design of a magnetic pole is considered in [10]. In both cases, different objective functions are associated with two independent methods of solving the considered forward problem.
Some existing methods for the inversion of multi-physics measurements in the area of oil exploration are based on requesting that geometrical structures identified by single-physics measurements are correlated (see e.g. [11]). Other methods employ experimental laws such as Archie's and Gassman's equations to relate different physical measurements among each other (see e.g. [12], [13]). Unfortunately, the aforementioned experimental laws contain various parameters that need to be properly adjusted, which is not always possible.
The more common idea of applying multi-objective optimization for solving IPs leads only to improving the method utilized for minimizing single ill conditioned misfit. A two-objective parameter identification by genetic algorithms can be found in [14]. The second additional criterion penalizes the small diversity in the populations of candidate solutions. Another approach used in [15] consists in combining two objective functions with an immunological algorithm. These objectives become fitnesses of individuals and T-cells, respectively.
The first approach intensively utilized in the sequel of this paper is related to the set of n physical processes ui(ω) ∈ Vi, i = 1, …, n, which depend on the same, unknown parameter . Typically, ω is a discrete representation of a distribution of some physical quantity (e.g. heat conductivity) on the dense domain of the forward problem. Vi are proper Sobolev spaces and Ai(ui(ω)) = 0 the relevant equations representing forward problems, where Ai : Vi ⟶ (Vi)′ is a family of differential operators from Vi to their conjugate spaces.
We are able to measure the state of all processes, which results in the vector , where are the sets of observations specific to each physic. Next, we introduce the misfit functionals f(d, u(ω)), such that . Each coordinate fi represents the particular physics i = 1, …, n.
The first multi-objective problem that represents the IPs associated with multiple physical processes dependent on the same parameter function consists of finding ω such that it minimizes all misfit functionals in the Pareto sense (see e.g. [16])
The above approach might be generalized to the case of many physical processes that depend on different parametric functions. Let us assume similarly that we have n physical processes ui(ωi) ∈ Vi, i = 1, …, n, which depend on a different, unknown parameter where is the admissible set. Their states can be measured and their measurements can be denoted as previously by . Now, we introduce the separate, metric space of features that represents the phenomenon to be recognized (e.g. oil deposit, tumor tissue) and the operators extracting features . The value Ci(ωi) represents the information upon the phenomenon under interest obtained from the ith physics. Their values may represent the characteristic function of the deposit region (e.g. the region of the oil occurrence).
In the sequel, we introduce a new incidence criterionthat penalizes the incoherency between the parameters assigned to the particular physics. In other words, the quantity fn+1(C1(ω1), …, Cn(ωn)) takes a small value if the parameters ω1, …, ωn represent a “similar” phenomenon to be searched. As the misfit operators fi, i = 1, …, n, the incidence criterion is strongly related to the particular IPs. In the simplest case n = 2, we may set , where stands for the metric function in the space of features.
The generalized multi-objective problem that represents the IPs associated with many physical processes that depend on the different parametric functions is formulated in a way similar to the problem (1):
We foresee that, due to the last objective fn+1, we will be able to seamlessly discriminate against those solutions obtained for the physical models that are too distant. More specifically, an artifact appearing in the particular ωi would be considered only if the other processes would yield enough evidence that the particular candidate solution is indeed promising. Thus, we expect tremendous reduction of generated artifacts and, in consequence, we expect severe reduction of multimodality of the problem (as many apparent extrema arise from interactions between artifacts and true extrema). Generally, we expect that solving many models jointly would yield information about inverse solutions that would be much broader and useful than that obtained if we considered each model separately. We hope that because of this additional information, the computational cost of solving (1) or (3) would not be larger (or, in fact, would even be possibly smaller than) the cost of minimizing each misfit separately.
Our proposed approach provides a rigorous mathematical framework that complements and generalizes existing methods for the joint inversion of multi-physical measurements. In particular, functions Ci defined above can be selected: (a) to impose some geometrical correlation, (b) to reproduce the behavior of certain experimental laws, or (c) to impose any other interrelationship of interest between different physical phenomena.
The generalized IP formulation (3) might be considered for solving oil deposit investigations on the base of the common inverting electrical resistivity and sound speed using independent measurements obtained by the electromagnetic and ultrasonic antennas.
The approach presented in Section 1.2 was briefly introduced in [17]. The current paper extends the first computational experiments contained in [18] and published in [19].
We apply the complex, multi-deme Hierarchic Memetic Strategy (HMS) [20] well suited for solving IPs for finding Pareto compromise solutions of the problem (1) (see Section 2) for the case of twin physics (n = 2). In Section 2.2, the special kind of the selection operator, a particular type of rank selection (cf. MOGA [21]) that prefers the coherent Pareto solutions is discussed.
In the sequel of the paper we perform the benchmark analysis (see Section 3) showing the exploratory power of the proposed strategy and the effective elimination of incoherent compromise solutions.
The last part of the paper (see Section 4) contains the solution of a real-world engineering problem of inverting magnetotelluric (MT) measurements (see [22]) in order to find oil deposits located under the Earth's surface. Two misfit functions are related to distinct frequencies of the electric and magnetic waves, for which the maximum sensitivity with respect to the impedance to be searched is expected. In this example, we hybridize for the first time HMS with a special kind of the local, convex optimization [23], [16] in order to increase the search accuracy.
Section snippets
Hierarchic Memetic Search
This section contains a short description of HMS, concentrating on its computational aspects. For the details on the system architecture and algorithms, we refer the reader to papers [20] and [22].
As a whole, HMS can be seen as a composition of a global optimization tool and an external direct problem solver. The latter is necessary for the evaluation of the objectives, the former seeks the global minima of the objectives. Naturally, an integration tier must be provided for an appropriate
Benchmark example
Computational benchmark studies presented in this section aim at illustrating and showing the advantages of the approach presented in Section 2.2. We compare performance of MO-HMS without and with RM on a two-objective multimodal benchmark problem imitating a real-world inverse problem with objectives induced by two physics models. The goal is to find regions in which both objectives have small and similar (high incidence) values, and to filter out artifacts that appear in one objective and are
Twin objective magnetotelluric data inversion
In this section we present an application of methods described in Section 2 in the solution of a real-world engineering problem of the inversion of magnetotelluric measurements. Note that here we use both the multi-objective selection operator defined in Section 2.2 and the local objective scalarization from Section 2.3.
Conclusions
We propose a new memetic strategy for solving multi-physics, complex inverse problems formulated as multi-objective global optimization ones. The objectives are misfits between the measured and simulated states of various governing processes. The multi-deme, tree-like structure of the population allows for both, intensive and relatively cheap exploration providing moderate accuracy results and a more accurate search of some regions of Pareto set in parallel. The special type of rank selection
Ewa Gajda-Zagórska received her M.Sc. degree (2009) in computer science at the Jagiellonian University in Kraków and a Ph.D. (2015) at the Adaptive Algorithms and Systems Group in the Department of Computer Science, AGH University of Science and Technology in Kraków. She was a recipient of Google Anita Borg Scholarship in 2013. Currently she is a postdoctoral researcher at IST Austria. Her research interests include evolutionary algorithms, physics-based simulations, multi-objective
References (47)
- et al.
Multiobjective hierarchic memetic solver for inverse parametric problems
- et al.
A hybrid method for inversion of 3D AC resistivity logging measurements
Appl. Soft Comput.
(2015) - et al.
Efficient gradient calculation of the Pareto optimal curve in multicriteria optimization
Struct. Multidiscip. Optim.
(2002) - et al.
First geoelectrical image of the subsurface of the Hontomín site (Spain) for CO2 geological storage: a magnetotelluric 2D characterization
Int. J. Greenh. Gas Control
(2013) - et al.
A secondary field based hp-finite element method for the simulation of magnetotelluric measurements
J. Comput. Sci.
(2015) - et al.
A self-adaptive goal-oriented hp-finite element method with electromagnetic applications. Part II: Electrodynamics
Comput. Methods Appl. Mech. Eng.
(2007) - et al.
Inversion of magnetotelluric measurements using multigoal oriented hp-adaptivity
Procedia Comput. Sci.
(2013) Inverse Problem Theory, Mathematics and Its Applications
(2005)- et al.
Regularization of Inverse Problems
(1996) - et al.
A hybrid method for inversion of 3D DC logging measurements
Nat. Comput.
(2015)
A problem in the optimal design of networks under transverse loading
Q. Appl. Math.
Multimodal function optimization with a niching genetic algorithm: a seismological example
Bull. Seismol. Soc. Am.
Damage detection with parallel genetic algorithms and operational modes
Struct. Health Monit.
A novel evolutionary algorithm for identifying multiple alternative solutions in model updating
Struct. Health Monit.
Multi-deme, twin adaptive strategy hp-HGS
Inverse Probl. Sci. Eng.
A novel workflow for the inverse QSPR problem using multiobjective optimization
J. Comput. Aided Mol. Des.
Comparison of multi-objective optimisation approaches for inverse magnetostatic problems
Int. J. Comput. Math. Electr. Electron. Eng.
Joint inversion: a structural approach
Inverse Probl.
Three-dimensional joint petrophysical inversion of electromagnetic and seismic data
Joint inversion of borehole electromagnetic and sonic measurements
Solving multimodal problems via multiobjective techniques with application to phase equilibrium detection
Fluid flow in hydrocyclones optimized through multi-objective genetic algorithms
Inverse Probl. Sci. Eng.
Nonlinear Multiobjective Optimization
Cited by (0)
Ewa Gajda-Zagórska received her M.Sc. degree (2009) in computer science at the Jagiellonian University in Kraków and a Ph.D. (2015) at the Adaptive Algorithms and Systems Group in the Department of Computer Science, AGH University of Science and Technology in Kraków. She was a recipient of Google Anita Borg Scholarship in 2013. Currently she is a postdoctoral researcher at IST Austria. Her research interests include evolutionary algorithms, physics-based simulations, multi-objective optimization and inverse problems.
Robert Schaefer is a full professor at the Department of Computer Science, Faculty of Computer Science, Electronics and Telecommunication, AGH University of Science and Technology. Author and co-author of about 180 books, papers and conference contributions, in particular author of the book Foundation of Genetic Global Optimization (Springer 2007). General chair of the PPSN 2010 Conference and Steering Committee Member of the PPSN Series. PC member and co-chair of more then 100 scientific conferences in computational sciences and artificial intelligence. Recent research areas: algorithms of oil and gas surveying, genetic and memetic algorithms, computing multi-agent systems, adaptive algorithms of solving forward and inverse problems for PDEs. Former research: modeling of the blood flow in arteries, modeling of nonlinear flow in porous media.
Maciej Smolka received his Ph.D. (2000) in the area of shape optimization at the Jagiellonian University in Kraków. Until 2012 he worked in the Chair of Optimization and Control Theory at the Jagiellonian University. Currently he is a member of the Adaptive Algorithms and Systems Group in the Department of Computer Science, AGH University of Science and Technology, Kraków. His current research interests include stochastic modeling of computational systems, application of evolutionary and hybrid algorithms in solving inverse problems and computational multi-agent systems.
David Pardo is a research professor at Ikerbasque, the University of the Basque Country UPV/EHU, and the Basque Center for Applied Mathematics (BCAM). He has published over 120 research articles and he has given over 200 presentations. In 2011, he was awarded as the best Spanish young researcher in Applied Mathematics by the Spanish Society of Applied Mathematics (SEMA). He leads several national and international research projects, as well as research contracts with national and international companies. He is now the PI of the research group on Mathematical Modeling, Simulation, and Industrial Applications (M2SI). His research interests include computational electromagnetics, petroleum-engineering applications (borehole simulations), adaptive finite-element and discontinuous Petrov-Galerkin methods, multigrid solvers, image restoration algorithms, and multiphysics and inverse problems.
Julen Álvarez-Aramberri completed his degree in physics at the University of Basque Country in 2007 after concluding his last year as Erasmus student at the University of Florence. Then, he studied a M.Sc. in Quantitative Finance (2007–2009) and a M.Sc. in ‘Mathematical Modelization, Statistics and Computation’ (2010) at the University of Basque Country. In 2015, he received the Ph.D. degree in applied mathematics in co-tutelage between the University of Basque Country and the University of Pau (INRIA, MAGIQUE3D group) with the work entitled hp-Adaptive Simulation and Inversion of the Magnetotelluric Measurements. Since January 2016, he is a Postdoctoral Visiting Fellow in Computational and Applied Mathematics at the Basque Center for Applied Mathematics (BCAM). His main research interests lie in the area of computational and applied mathematics. In particular, in the efficient implementation of numerical schemes, methods, and tools to efficiently model, simulate and interpret real word problems. Specifically, in solving direct and inverse problems arising in the application of the magnetotelluric technique, a method employed to retry the resistivity distribution, and hence to obtain an image of the Earth's subsurface.
- ☆
The work presented in this paper has been partially supported by Polish National Science Centre grant no. DEC-2015/17/B/ST6/01867 and by the AGH grant no. 11.11.230.124; Ewa Gajda-Zagórska was funded by Polish National Science Centre grant no. DEC-2012/05/N/ST6/03433; D. Pardo and J. Álvarez-Aramberri were partially funded by the RISE Horizon 2020 European Project GEAGAM (644602), the Project of the Spanish Ministry of Economy and Competitiveness with reference MTM2013-40824-P, the BCAM Severo Ochoa accreditation of excellence SEV-2013-0323, and the Basque Government through the BERC 2014–2017 program and the Consolidated Research Group grant IT649-13 on “Mathematical Modeling, Simulation, and Industrial Applications (M2SI)”.