Case study: Nexia solutions

Modelling and simulation in the nuclear industry

This is an extract from an article that appeared in tce (the chemical engineer) in September 2006. [Download full article pdf overview ]

Nexia Solutions uses Process Systems Enterprise's gPROMS advanced modelling package for modelling of complex nuclear and chemical systems. Vicky Ashley, Richard Jarvis and Scott Owens describe why and how they use the package, and the key role of modelling within Nexia Solutions.

Nexia Solutions

Nexia Solutions Limited formed from the Research and Technology (R&T) subsidiary within the British Nuclear Fuels plc (BNFL) group of companies.

The organisation has over 30 years of experience and specialised knowledge of nuclear chemistry and chemical engineering processes, in application areas such as waste management, fuel reprocessing, asset maintenance and treatment of radioactive effluent.

Nexia's main function is to provide R&T support to nuclear sites to ensure their continued safe operation.

We are also currently gearing up to play a major role in a future national nuclear laboratory.

The role of our department is to underpin these services by providing model-generated data for decision support, using a variety of software tools.

As a supplier to the nuclear industry, we face particular challenges. We have to deal with complex chemistry that can involve two-thirds of the periodic table. Our customers' plants cannot be accessed or modified easily, and it is virtually impossible to add instrumentation. In addition to this, there is a critical need for safety in everything we do.

SIXEP gPROMS flowsheet

The authors

The role of modelling in Nexia Solutions

As a provider of services in this complex technology area it is essential that we can justify decisions quantitatively to our clients.

Modelling enables us to do this, in a reproducible way. Equally important, it allows us to capture knowledge, deploy it to add value and at the same time build on that knowledge to develop an increasingly accurate predictive capability.

To model effectively, we first need to understand the science which underpins our customers' plants and processes and capture this knowledge in model form.

Once this is done, the models are then used to provide support to front-end and detailed design, in order to improve design, safety, operations and economics by investigation and resolution of plant problems, as well as to develop and scale-up new processes.

In particular, modelling is used to minimise risk by removing uncertainties. It also provides a Quality Assurance audit trail, to demonstrate that we have done the analysis, and to show how we came to the conclusions that we did. This is essential in the nuclear industry.

gPROMS is our key tool for process modelling. We adopted the package in 2005 as a replacement for our existing APM and flowsheeting software, which we felt was not developing in the way that we wanted and had some modelling and robustness limitations.

Our year-long evaluation involved testing various modelling tools on the market against a rigorous list of "ideal software" criteria. We also investigated the working relationship with the suppliers, as this is important to us.

gPROMS and PSE came out highly in all areas; there were some limitations, but - crucially - no 'show stoppers'.

How and why do we use gPROMS?

We use gPROMS as a modelling and solution framework. The software is used primarily by two groups: process modellers and chemical modellers; each has different requirements, but both need powerful modelling and solution tools that are easy to understand and work with.

Process modellers typically work with a process flowsheet representation which sees the process in chemical engineering 'unit-and-stream' terms. Chemical modellers use gPROMS as a dynamic solver, developing detailed chemistry models that are delivered back into the chemical engineering environment to provide increasingly-accurate predictive models.

In the past we tried mathematical solver software for the chemistry side, but this led to complex and uncheckable models for large sets of dependent reactions. It was much more difficult to access information and maintain models than it is in gPROMS. Another big advantage of our current setup is that both the process and chemical modelling groups can work with a common tool.

A typical 'project lifecycle' involves developing and QAing models, performing sensitivity analyses, performing parameter estimations to adjust model parameters to laboratory or pilot data, and analysing model results. Only when this has been done do we deploy models and scale up results. An essential requirement was that the environment could cope with all these activities. Because of the type of applications we have, most of our models need to be custom built.

Another key aspect is the linking of modelling with experimental work and pilot plant trial data. Because of the costs and safety implications of running radioactive experiments and pilot plant trials, we build existing knowledge into models which are then used to reduce the number of experiments required to a safe minimum.

We also use models to interpret experimental data to provide accurate parameter values and confidence information, which is important in risk management. Once we have sufficiently accurate parameters, we can use this information to reliably scale-up the process. In the future we will apply gPROMS' formal model-based experiment design techniques to design optimal subsequent experiments, minimising the number of factors to be considered and thus the experiment time, costs and materials requirements.

Example

A typical application is the SIXEP plant , an ion exchange process designed and commissioned in the 1970s to remove radioactivity from a particular effluent streams. In recent years, the decommissioning of plants means that there is a strong commercial motivation to use SIXEP to treat additional effluents.

However before our customer does this they need to prove to themselves and the Environmental Agency that the process will still perform effectively and reliably. Though the original engineering was very sound, there is no obvious way to extrapolate the original data to validate performance for new feed compositions and process conditions.

This is where predictive modelling can help. Over the years we have built a high level chemical engineering description coupled with increasingly accurate chemistry models. The current gPROMS model now embodies the SIXEP engineering, physics and chemistry (down to ion exchange levels) in a mixture of first-principles and empirical models backed up with experimental data. Before any recommended changes to the plant, we go through a rigorous modelling procedure, which includes a whole raft of experiments and QA to test model validity.

The customer now considers this model as an essential requirement for their plant operations. The components of the model have since been applied to new ways of deploying ion exchange technology and in the evaluation of new materials.

Impressions

Our major impression is that gPROMS is numerically very powerful, and provides all the process modelling capability that we need. This was our number one requirement. It is also well-structured to the way that an organisation such as ours works, with a lot of interaction between different technology groups and rigorous QA procedures.

Important features here are the model management capabilities, the 'openness' of the gPROMS models, and the powerful audit facilities. We also like the gO:Run facility, which allows gPROMS models to run in 'execution mode' behind an Excel interface. This means that we can deliver packaged models (encrypted for confidentiality) for end users to carry out their own investigations. We expect this to increase in importance as we deal with a wider range of clients in the future.

A seemingly minor but important aspect for us is that we can run the software on our own server, which means that gPROMS can be used on any PC. Updates are downloaded from the web straight onto the server, where they are immediately available to all users. We can also execute applications in batch mode if necessary.

Naturally we have a few grumbles. For a start, we believe that parameter estimation facilities in gPROMS are too geared towards expert users. Quite often, all chemists want to do is a quick least-squares fit to data; there needs to be more default options for the infrequent user. We have also had some issues to do with spoecies handling in arrays and the ability to execute runs interactively from within the gPROMS ModelBuilder environment rather than via the Excel interface. PSE is addressing these items.

One of the things we are very happy with is the relationship with PSE, who provide us with excellent support and regular modelling assistance. Also important to us is the fact that, as a key user in a key sector, we are members of the PSE Strategic Advisory Council. This gives us a strong say in development priorities, and confidence that our requirements will be taken into account.

The future?

The potential impact of modelling on R&T is something that we will exploit much more in the future, using the full power of gPROMS for model-based data analysis and experiment design. PSE are working with us to introduce these capabilities.