Edouard Demaldent | CEA LIST
24th September 2025
Today’s interview brings us right at the core of the future of CIVA software, with Edouard Demaldent, Head of the Simulation, Modeling and Analysis Laboratory within the Digital Instrumentation Department at CEA LIST.
Could you say a few words about the activities of the department, but especially about those of the laboratory you manage?
Our department draws on dual expertise in instrumentation and modeling, which it uses to analyze and innovate inspection processes. Designed to bridge the gap between these different areas of expertise, the CIVA software platform we are developing was born from a desire to capitalize on and, above all, share this expertise. Accelerated by the creation of EXTENDE about fifteen years ago, which coincided with my arrival at the CEA, this commitment has remained intact and even seems to me to be reinforced today.
It naturally draws on the activity of the simulation laboratory, which ranges from academic research to software integration, with an increase in maturity of our models which takes place continuously alongside our partners in advanced studies.
A second dimension of the laboratory’s activity relates to the multi-physical and multi-technical characteristics of the platform, which addresses the simulation of the main inspection methods, with semi-analytical, numerical, or even hybrid calculation solutions.
Finally, we align ourselves with current challenges and uses, which by definition vary over time. When I started, adding physical contributions to our models was a priority. Then, the ability to conduct parametric sensitivity studies with statistical analysis tools was prioritized. Today, a major challenge is producing labeled synthetic data with a realistic rendering, for training purposes as well as for training automatic diagnosis algorithms. With this in mind, the close collaboration of our simulation laboratory with other entities in the department, closer to instrumentation and diagnostics, is critical.
A new version of CIVA software is being released in September 2025. One of the major new features concerns the UT module with the availability of a “Finite Elements” perspective. Could you describe the motivations and history of the integration of finite elements within the CIVA platform (UT but also other methods)?
CIVA’s standard modules rely on semi-analytical approaches to access simulation studies that include scans and variations on a simple PC. Various hybridizations have extended their scope over the years. This is the case for the ultrasound module, which includes a hybrid defect response mode based on the Finite Element Method, functional since 2017 and which will continue to evolve.
This dynamic is constrained in the sense that we are aware that too many calculation options raise questions from the user experience point of view. On the other hand, hybridization does not free us from all the operational limits related to field calculation in the defect-free specimen, such as taking into account Rayleigh and head waves or even certain critical angles. Finally, we believe that it is healthy to be able to compare optimized hybrid semi-analytical simulations to a reference solution on a few critical configurations in order to reinforce the conclusions of a study.
These different reasons led us to develop a new complete FEM module, complete in the sense that the propagation of the ultrasonic wave is simulated throughout the part, which includes all ultrasonic modes, edge effects and the flaw. An ambitious but achievable objective was to make possible a sensitivity study in 2D (or low frequency 3D) configurations, and to make accessible the evaluation of a few shots on complex 3D configurations at a few MHz.
We knew that this objective was compatible with the FEM strategy developed in the laboratory over the past ten years because it is relatively light in terms of calculation time (compared to other FEM solutions) and above all has a reduced memory footprint. Available since 2017 in hybrid defect response mode and 2021 in the integrated Structural Health Monitoring (SHM) module, the challenge for us was to bring this FEM solution to the specificities and standards of the ultrasound module.
What are the specificities of the approach used for integrating finite elements in CIVA UT? What are its advantages, but perhaps also its limitations?
The foundations of this finite elements module date back to the 2000s with the rise of the transient Spectral Element Method (SEM) for ultrasonic applications. It is a variant of the FEM with high-order polynomials that drastically reduce numerical dispersion phenomena, i.e. the increase in numerical error with simulation time and the distance traveled by the ultrasonic wave, which is a critical issue for our inspection applications. For a high-order method, the main strength of the SEM lies in its performance in terms of computation time and memory footprint. On the other hand, reaching its highest level of performance requires the construction of a regular mesh of quadrilaterals (hexahedra in 3D), which does not correspond to the standard of meshing algorithms, and thus makes it a difficult method to integrate into NDT software.
We took our first steps with a thesis and then a post-doc in the late 2000s alongside our academic partner POEMS (UMR 7231 CNRS-INRIA-ENSTA). After a complete overhaul of these first models, we carried out a first maturation stage with a domain decomposition strategy to ensure the stability of the simulations, optimize the distribution of unit calculations by sub-domain, and above all remain in control of an imposed over-division into quadrilaterals/hexahedra. This stage notably resulted in the development of original design tools, for example to draw the grid around complex defects in CIVA. Largely supported by the department’s software engineering laboratory, this phase constitutes the main performance base of CIVA’s FEM solution.
The next phase consisted in extending the functionalities of the FEM solver and the complexity of the processed digital scenes (attenuation, thin layers, slightly inhomogeneous media) without deviating from this performance base. This constraint feeds cycles of monitoring, research, mathematical analysis, design, implementation, and evaluation, which today constitute a significant part of our scientific activity.
The third phase was the implementation and evaluation of these solutions with our industrial partners, particularly for the applications of EDF, who has been supporting us for many years in the study of advanced methods. This naturally led us to an initial model in CIVA, shared internally from the beginning of 2023 but still far from being marketable.
Over the past two years, an internal project has been fully dedicated to this FEM module and the user experience, adding features and fixing bugs. Among the many advances made, I can mention the automatic selection of windows in space and time of the meshed area of interest, the possibility to define defects in the field calculation module, the introduction of mixed meshes of triangles and quadrilaterals (extruded for 3D simulations) on a CAD description of the inspection scene.
Today, we are proud of the progress we have made and confident in the contribution of this module to CIVA users. However, we remain well aware of the limits we had to set ourselves, both in terms of performance and genericity, and remain committed to the next steps to come. For example, despite the progress made in mesh libraries, sometimes a few localized meshes in the digital scene penalize calculation times. To overcome this risk, we are developing and evaluating a technique for extracting and then specifically processing these meshes, which we hope to make available within 18 months.
Beyond the “commercial” version of CIVA, actively used by more than 350 entities worldwide, the CIVA platform is also the subject of numerous developments within the framework of more specific industrial projects. Could you mention some of these themes?
Developing a new CIVA module is an exciting adventure for NDT simulation experts like us, as it means bridging the gap between the capabilities of a scientific computing code (FEM or other), the result of our research, and the software features expected by an NDT professional. Regardless of the level of maturity reached, it is always surprising to see the leap we still have to make to meet new needs. This feedback, in turn, feeds into our roadmap, sometimes motivates the launch of a new research topic, and forces us to constantly question our solutions.
Most often, a specific combination of the modeling technology building blocks we have or are studying meets the demands of our partners. Regarding FEM for ultrasonic inspection, this may involve a particular 3D geometry meshing strategy using different calculation techniques, such as taking into account inter-ply spaces in curved composites or interfaces between aggregates and cement for concrete.
In this case, we propose to adapt CIVA to include specific calculation options, sometimes in the form of a dedicated proprietary plugin module. This format allows us to arrive at a first deliverable version evaluated on a pre-established specification within a few months, which then evolves according to our partner’s work program, and which we can maintain. For us, this is an excellent way to increase maturity both in terms of simulation technologies being tested and in terms of the purpose of simulation for new inspection processes.
Could you describe the main perspectives for future versions of CIVA, whether around finite elements models or other aspects?
We are not lacking in imagination or enthusiasm when it comes to discussing the future. Among the most advanced topics involving finite elements for acoustics, but on different technical bases than those mentioned above, we are developing a simulation module taking into account a microstructure in small 3D samples. A first objective is to deduce from these predictions, made at the microscopic scale, homogenized material properties for our inspection simulations at the macroscopic scale. The challenge is to strengthen our knowledge and therefore confidence in the description of the digital scene. A second challenge is to open the CIVA platform to new applications of ultrasonic characterization, by bulk waves then by guided waves. On the theme of guided waves, we are thinking about how to converge our inspection and structural health monitoring modules (GWT and SHM).
On the electromagnetic inspection side, we are following a similar strategy to acoustics, strengthening CIVA’s standard semi-analytical models with new digital solutions. For example, the module dedicated to steam generator tube inspection is enhanced in the new version. Above all, we are working on a future module compatible with 3D CAD geometries. The ability to ensure an efficient heating coverage for induction thermography of aeronautical parts is particularly motivating this approach. At a less advanced stage, we are preparing a 3D thermal diffusion model adapted to surface temperature gradients and are working on the design of an electromagnetic characterization module.
In X-ray radiography, we are continuing to enrich the computing cores to broaden the range of interactions simulated in CIVA. This will be reflected in 2025 with the arrival of X-ray fluorescence or the inclusion of “Bremsstrahlung” continuous radiation braking, while photon counting detectors will soon be available. On the other hand, we are undertaking a project to modernize our tomographic reconstruction models. One challenge for us is to address the problem of reconstructing large parts in scattered views, which pulls forward other innovations under study, such as the selection of the best views with regard to inspection and interfacing with a robotic experimental cell.
Finally, I would like to highlight the arrival of a first augmented imaging concept at the interface between the simulation modules and the recent data science (DS) module. The insertion of a simulated virtual flaw into defect-free experimental imaging makes it possible to obtain a more realistic texture and therefore a more realistic rendering. Available at this stage for X-ray radiography (RT), we intend to expand this perimeter in the coming years, particularly for ultrasound imaging.