Tracking down leaks with AI

Many pipes are like “ticking time bombs underground”. This is how PipePredict describes the problem. If a leak occurs – for example in a water pipe – the predetermined breaking point often grows exponentially over years until it breaks. “Then so much water has accumulated that an entire street can be washed out and sink,” explains co-founder Valerie Fehst. But it is not only such infrastructure damage that concerns the physicist and her two business partners Christopher Dörner and Tri-Duc Nghiem, but above all the enormous loss of water due to undetected leaks in the supply system for water and also for district heating.

For the benefit of society

“We are in the fortunate position in Germany that we have hardly any water overuse, but we know: Worldwide, there will be less and less drinking water for more and more people,” Fehst emphasises. The new PipePredict technology is intended to contribute to resource conservation and sustainability and create added value for society as a whole. The idea is based on an AI algorithm that Fehst developed as part of her master's thesis in the Department of Physics at TU Darmstadt. It makes it possible to automatically analyse vibration patterns that occur when a fluid – water, gas or even oil – flows through a pipe under pressure. "As soon as there is a leak in the pipe, the pattern changes. And this is our starting point,” she explains. The result is a software-as-a-service solution that monitors even hard-to-reach pipes in real time, regardless of the material, locates even small leaks early and precisely, and allows for predictive maintenance of supply networks.

Pilot customers in Germany and France

PipePredict works with the sensors that are already installed in the supply network. From their data, the team creates a digital image of the network, analyses the current state and then builds the necessary specifications together with the customers. This is not only exciting for private and public water and energy suppliers, but also for industrial production and the operation of oil and gas pipelines. The focus of acquisition is currently still on water and district heating service providers. In this area, the Darmstadt-based start-up is already underway with paid pilot projects in Germany and France. A pilot with a private energy supplier is currently being developed into an integrated software-as-a-service for regular operation. Such a conversion brings not only technical challenges, for example with regard to the strict security standards for critical infrastructures, but also entrepreneurial ones: “We now have to convince not just one, but a large number of stakeholders,” Fehst reports.

In the meantime, the PipePredict team has provided the “proof of concept” for its product in many places, the business model and the structures are in place. More pilot projects are to follow – also to further optimise the software under real conditions. On its way so far, the team has received a lot of support, for which Fehst is “super grateful”: from HIGHEST around the start-up process and the application for EXIST funding, from the TU Darmstadt, with which the start-up has meanwhile implemented a bachelor's and master's thesis, and from the Darmstadt start-up centre HUB 31 and the Lab³ association, which provide the infrastructure for further research and development work at affordable costs. “In addition, many investors believed in us early on,” says Fehst. They include Powercloud founder Marco Beicht and the Leipzig-based venture capitalist Smart Infrastructure Ventures.

Team spirit

Above all, PipePredict's success stands or falls with the team. “I didn't have the courage to found the company on my own,” says Fehst. The current head of technology knew that she was very well positioned in her sub-field, but needed know-how in computer science and business administration as well. Industrial engineer Christopher Dörner and Tri-Duc Nghiem, an expert in Natural Language Processing (NLP), were able to inspire her to “join in the uncertainty of a start-up”. In view of the increasing demand, the three now need more reinforcement, have to build the structures for a growing team to work together and a common corporate culture. “We can't offer a huge salary like in industry,” Fehst says. “What we offer is freedom – to think and to shape one's own work and life.”