IBA invests in Machine Learning to predict defects

The IoT allows us to perform predictive maintenance on our proton therapy devices.

IBA uses IoT to enable predictive maintenance for its devices
Xylos designed the solution’s Azure infrastructure

Every downtime of an IBA proton therapy system has a significant impact on patient treatments. If the downtime is significant enough, the patient will need to be rescheduled to a later date, with a high risk that the cancer will grow further in the meantime. Fortunately, most downtime can be prevented with adequate monitoring and scheduled maintenances. Still, IBA has the ambition to go one step further and predict when their devices would malfunction so they can replace the worn parts outside the patient treatment hours.

To achieve this, Xylos helped IBA architect an Azure cloud IoT infrastructure to capture the machine data from all the company’s proton therapy systems all over the world. This architecture will in a next phase allow for Machine Learning software analysis to predict potential issues before they arrive.

The customer and IBA would benefit from keeping the lifetime and performance of their systems as high as possible. Linking the devices to the Internet of Things makes it possible for IBA to monitor them more efficiently and even predict potential defects. To achieve this, IBA turned to Xylos’ IoT experts for help.

Predictive maintenance

“Thanks to IoT, big data and artificial intelligence, we can now go for remote monitoring and predictive maintenance”, says Kurt Verduyckt, Digital Solutions Manager. “This enables us to reduce the total cost for service and operations while also improving our customer service. Currently on the bigger centers we still have up to 12 IBA engineers on site to keep the complex system operational at all times and reach our high uptime agreements. We don’t only sell a product, we also sell a service! Hospitals, oncologists and patients benefit from this as well: for every hour the device is out of order, no therapies can be started. In other words: this doesn’t only have financial, but also clinical consequences.”

A proton therapy system is very complex and has a lot of different sub components that all generate a lot of data. Xylos helped linking all these data sources to the Internet of Things via Azure IoT Edge, that also makes sure to anonymize the data before it’s sent to Azure via Azure IoT Hub. Once there the data is distributed to different destinations. One of them is Splunk where the data can easily be queried. Another one is the Azure Data Lake where the information is kept for further processing. In a later phase, Azure Machine Learning technology will be able to process this data and compare them to previous values or even to similar systems on different locations. This way, the software could detect problems early on, which allows IBA to intervene swiftly or even plan a preventive maintenance. This entire data flow was designed to allow real time processing.

  • 12 engineers maintain the systems on site today
  • 4 different kind of data types collected by Azure
  • +10.000 different machine signals captured by the IoT solution

Xylos designs IoT service infrastructure

Thanks to Xylos, IBA can now concentrate on developing its predictive maintenance service. Kurt: “Xylos made sure the right infrastructure is set up in the Azure Cloud, with solutions like Azure IoT Edge to capture the data, Azure data Lake to store the data and Azure Data Bricks to analyze the data. We now have a clear blueprint for the next steps and don’t have to worry about anything. We can develop our own functionalities and add them to the system. We’re evolving towards a digital product team that covers infrastructure and software and we can work with Xylos in an agile way.”

The IBA development team came up with the idea of using Microsoft Azure because it would reduce the efforts of deploying the solution to all systems over the world. On each site you now only need to install the relatively basic IoT edge connector that you can even update remotely with new functionality as needed. The real data monitoring and analysis application is installed only once in the cloud, which drastically speeds up the deployment of new features for all systems at once. As IBA continues to sell more proton therapy systems, this cloud computing model is also a lot easier to scale. Finally, Microsoft Azure already offers countless ‘libraries’ for the development of IoT and Machine Learning solutions.

Flawless preparation by Xylos

IBA trusted Xylos with the project because the company has experience with similar projects. “We asked Xylos to design and create the infrastructure. They’ve got plenty of experience with Azure and expertise when it comes to infrastructure and development. They’re also very pragmatic and they were incredibly flexible when it came to the project duration”, Kurt Verduyckt says.

IBA also asked Xylos to write a governance file before the start of the project. To do this, Xylos analysed which Azure technologies IBA would need to set up the service, manage the solution and integrate it into the current architecture, and what the project would cost.

Next steps

Today, most of the IBA proton therapy systems are connected. “We are now working on a first predictive maintenance model using state of the art machine learning techniques and the data we now have readily available.”

“Xylos made sure the right infrastructure is set up in the Azure Cloud, with solutions like Azure IoT Edge to capture the data, Azure data Lake to store the data and Azure Data Bricks to analyze the data.”

Kurt Verduyckt, Digital Solutions Manager