Bright Renewables - From Webinar to an Actual Predictive Maintenance Project

Under the header “FastForward”, itsme supports product improvement and process acceleration, including by informing customers about time-saving solutions and smart technologies. For example, in spring 2023, a webinar on predictive maintenance was organised in collaboration with KIM Plus Delta, consultants for asset and maintenance management. Bright Renewables’ participation resulted in an actual predictive maintenance project for Bright's green gas plants.

Bright Renewables in Enschede, part of the HoSt Group, supplies biogas upgrade systems. These convert biogas - derived from, among other things, fermenting food waste, agricultural products, animal manure or sewage sludge - into green gas by increasing the methane content. HoSt is responsible for constructing the fermentation plants. Biogas consists 50/50 of methane (the main component of natural gas) and CO2, plus some H2S (famous for its rotten-egg smell) and volatile organic compounds. Upgrading involves the separation of the different fractions, including membrane filtration and carbon vessels that capture H2S.

More Efficient Use of Manpower

Predictive maintenance (PdM) was already on Bright’s agenda, says process engineer Vincent Kroeze. “We collected a lot of data, but we weren’t looking at it in a smart way. We wanted to do more with that data. That’s why I participated in the webinar, to get a better understanding of the basics of PdM and explore the possibilities. That’s how I got in touch with Johan Koopman.” Koopman is an asset management consultant at KIM Plus Delta and a team leader at Bredius Datalab, the independent branch for data science for PdM. “Vincent immediately raised a practical question about predicting carbon vessel bleeding.”

If the carbon is saturated with H2S, a vessel needs to be changed, Kroeze explains. “A single person needs an entire day on site to do this. However, the number of actions, like flushing and emptying, is limited, so our service staff member mainly spends the day waiting. We wanted to combine this with regular maintenance and, before that, be able to predict when such a carbon change was imminent, so we could deploy our staff much more efficiently. The reason behind this being that although the demand for biogas plants has grown tremendously, our service organisation is not growing in a linear fashion because we are unable to recruit enough technicians.”

“Bright already had a good business case, a lot of domain knowledge and historical data.”

Actually, the bleeding of carbon vessels was not entirely a black box, Kroeze explains. “We are acquiring a better understanding of it. For example, it depends on the biogas quality and is very site-dependent in terms of how the biofermenter is operated. Dutch water boards, for example, have a continuous supply of sludge as feed from water treatment, with a stable H2S content, and generally good control over the process as well. We can already reasonably predict bleeding there, but that is not the case with other fermenters.” So he wanted to make the move to PdM with KIM Plus Delta.

As far as Koopman is concerned, they hit the ground running. “Multiple parties want to get started with PdM, but we often see entry problems, particularly in three areas. Bright, for the most part, had already figured things out. First, they had a good business case. From what we see in the market, there is not much thought given to this yet. It’s more like: “Oh, we have to do something with data.” Secondly, Vincent already knew a lot about the workings of carbon vessels, with enough domain knowledge to translate it into a data problem. Thirdly, the cherry on the cake, was the historical data of carbon vessel bleeding, which we could use to train a model.”

Inclusion of Purchasing Material in Forecast

Koopman tackled the issue from his engineering background. “With data scientists, I often find that they just take a bunch of data and run a model on it. I think domain knowledge is extremely important because it provides a context to the model to be set up. So my first question was: what are the comparisons of this bleeding process?” Based on this, Koopman and Kroeze started looking for functions with predictive value and associated variables, such as concentrations of biogas compounds, pressures, flows and temperatures. Due to the cumulative nature of the build-up to bleeding, they ended up with integrals and a linear regression model with non-linear components. In the meantime, tests have been carried out with a first model, reports Koopman. “The first results are encouraging, but do show dispersion; there is still room for improvement.” Kroeze is already pleased with what has been achieved. “If we know two weeks in advance about the upcoming bleeding, we have way more flexibility in our staff’s schedules. A spread of a few days would work very well for us.”

“If we know two weeks in advance about the upcoming bleeding, we can schedule maintenance accordingly.”

itsme’s account manager Bas Veneman also sees opportunities. For new-build facilities within the HoSt Group - including a HoSt site in Poland - itsme mainly supplies components to suppliers building modules. itsme itself mainly supplies spare parts to HoSt for the service organisation. “You could add purchasing the necessary materials to the forecast of bleeding and related maintenance. If bleeding is forecast in an x number of weeks, the purchase order can already be issued to itsme. We can then deliver the materials exactly on time, for example at the relevant site or at the technician’s home address. We are already in talks with HoSt about the list of materials and how we can also properly handle the long tail.”

Kroeze is happy. “The faster our company grows (HoSt now has 500 employees, ed.), the more this is needed. Ten years ago, everyone knew each other; now it’s impossible to know everything about everything. Three years ago, we still did a lot in MSExcel. In the meantime, we have implemented an ERP system and are digitizing a lot.” HoSt can also no longer keep stock of all materials itself, Veneman adds. “It’s too expensive. That said, we do have that stock and we can stock materials specifically for you.” It is therefore not an overtly bold prediction to say that the relationship between HoSt and itsme will grow even further.

Accessible and Broadly Designed

And all this as a spin-off from a webinar. Koopman: “We often connect like this, but rarely as hands-on as we did here.” Veneman: “It’s great that we as itsme have been able to make the connection.” Kroeze: “We were obviously on a quest and we had a specific question; but we were unfamiliar with the field. This webinar was accessible and wide-ranging. It was also good for us to gain insight into Bredius’ PdM roadmap.”

Check out our other success stories:

Stay up to date with the latest news:

Why not take a look at these trainings we offer: