Customer Story

Mitigating the environmental fallout of a nuclear era

An R&D scientist looks to design of experiments and other statistical methods to speed important research in nuclear waste conditioning

CEA Marcoule 

ChallengeDevelop a safer way to dispose of hazardous radioactive waste.
SolutionUse predictive models, design of experiments and other statistical applications in JMP to perform robust analyses that support new glass formulations and provide insight into vitrification processes and glass material performance.
ResultsBy adopting more efficacious research methods, scientists at CEA are able to process more information faster and at lower cost to the institution. Efforts saved via streamlined experimentation can be reallocated to support other relevant initiatives.

With the effects of global climate change growing increasingly tangible every day, the need to reduce greenhouse gas emissions and develop better mechanisms for renewable and low-carbon energies on a broad scale has become quite urgent indeed. In Europe, the government of France is leading the way with its Alternative Energies and Atomic Energy Commission (CEA). This 16,000-person research juggernaut, which in March 2017 was rated by Reuters as Europe's most innovative public research institution (and No. 2 in the world), is the institutional embodiment of “putting your money where your mouth is.”

Public institutions play a crucial role in expanding the bounds of nuclear science and technology

“There are so many challenges in energy technology today. That is why our job as scientists is so exciting,” says Damien Perret, R&D scientist specializing in nuclear waste conditioning at CEA. Perret is based at CEA’s Marcoule site, a research center devoted to preparing for a low-carbon energy future in the era of nuclear power. R&D activities at the facility include nuclear industry optimization, recycling of reusable materials in spent fuel (uranium and plutonium), dismantling of nuclear facilities and radioactive waste management.

The latter is Perret’s domain – a field that dates back to the mid-1960s in France when the CEA first began using borosilicate glass as a containment matrix for waste from spent nuclear fuel. Today, the transformation of nuclear matter into glass, or vitrification, is the internationally recognized standard with which to minimize the environmental repercussions of waste disposal. “Vitrification is the best compromise in terms of containment – chemical durability, thermal stability and resistance to irradiation – technological feasibility and cost (via the volume reduction factor),” Perret says. “For now more than 40 years, French scientists, engineers and nuclear experts from CEA and Orano have been dealing with vitrification in order to provide a comprehensive solution to high-level waste management issues.”


“With the same software, I can connect to our internal database, extract selected data by using filters, combine data tables, clean up my data, visualize it and apply powerful statistical techniques to analyze it.”

– Damien Perret, PhD, R&D Scientist

Design of experiments is a highly cost-effective method to speed critical research

Perret seeks to design optimized glass formulations that are specially adapted to waste characteristics. Among other things, an optimal formulation must meet the technological constraints of the manufacturing process and produce adequate final outcomes with respect to disposal requirements. “This job,” he says, “involves carrying out experimental studies related to glass material elaboration and characterization, computing the data with an internal database and analyzing this data to build predictive models.”

In material formulation, researchers seek to understand how changing the content of one particular compound might affect the final properties of a material. Design of experiments (DOE) enables them to adjust all factors at the same time, rather than running a multitude of single-factor tests. “A crucial point is related to the number of experiments you intend to run in a DOE. In our field, one single run involves many tasks related to glass elaboration and characterization steps. But the final cost for each run is significant.” DOE helps define with as much accuracy as possible the exact number of runs necessary to effect the desired outcome.

The cost of experimentation in the field of nuclear glass formulation is particularly high. That’s one of the reasons, Perret says, that it is so important to make use of decades of important historical data. “That means first doing a statistical analysis of existing data and second, getting the maximum information possible for future experimentation with a minimum number of runs. Cost reduction vis-à-vis reduction in test runs is typically the objective and benefit of using a DOE strategy,” he says. “In fact, we are facing many questions requiring the use of statistics, for example to develop the best statistical methodology for defect analysis, or to analyze results coming from a chemical analysis and compare these results with theoretical composition data.

“Today at CEA, scientists may use several different tools to visualize their data, design experiments or perform data preparation and statistical analysis. But I am always very enthusiastic to try to convince them that all of this can be done with one single software.”

JMP® is an all-in-one tool for data analysis

In JMP, Perret has found that single software. “I really enjoy the fact that with the same software, I can connect to our internal database, extract selected data by using filters, combine data tables, clean up my data, visualize it and apply powerful statistical techniques to analyze it,” he says. “I am a big fan of the Graph Builder platform, which is something I have never seen in other software packages. To me, this platform is extremely innovative.”

Perret says the Mixture Design platform, which supports multifactor compound experiments, is another functionality in JMP that he uses nearly every day. By defining a set of linear inequality constraints, researchers can limit the geometry of a mixture factor space. In Perret’s work on vitrification, he says the Mixture Design platform “helps us to define glass composition domains and to build property-composition predictive models.” Others on Perret’s team use JMP coding to write their own scripts. With the JMP scripting language JSL, Perret says users can “almost do everything.”

A highly engaged JMP® user community encourages statistical learning and knowledge transfer

Though Perret and many of his colleagues work with JMP on a regular basis, he says that they still use only a fraction of the software’s wide-ranging applications. Fortunately, an active network of JMP users around the world helps to share knowledge, allowing users with less statistical fluency like Perret to learn from some of the world’s foremost statisticians. “I am originally a chemist with almost no education in statistics, but I developed an honest background knowledge in stats just by using JMP,” Perret says. “Obviously the process of acquiring knowledge is infinite… but we have the JMP community of users and developers, which is a great help to progress.”

Lastly, Perret cites the reflexive engagement of JMP’s development team: “It is easy to see that the primary goal of JMP developers is to facilitate users’ lives by helping us to take the best decisions when doing statistics with JMP.” The responsiveness of the software to meet new challenges is one of the primary reasons researchers like Perret are willing to experiment with new methodologies and processes. For example, he says, “we have just started with a new project in which we are seeking innovative methods to create predictive models that are even stronger than what we did in the past. This involves creating a new database, new types of models and scripting. This is a very challenging new project for which we plan to use JMP.”


CEA would like to acknowledge their partners Orano and EDF for their participation in this project.

The results illustrated in this article are specific to the particular situations, business models, data input and computing environments described herein. Each SAS customer’s experience is unique, based on business and technical variables, and all statements must be considered nontypical. Actual savings, results and performance characteristics will vary depending on individual customer configurations and conditions. SAS does not guarantee or represent that every customer will achieve similar results. The only warranties for SAS products and services are those that are set forth in the express warranty statements in the written agreement for such products and services. Nothing herein should be construed as constituting an additional warranty. Customers have shared their successes with SAS as part of an agreed-upon contractual exchange or project success summarization following a successful implementation of SAS software.