Sensor on sensor - predictive maintenance of (power) electronics

Did you know it's possible to determine when an electronics component is about to break down?

RISE is a Swedish Institute and we have a long history of working in electronics packaging, integration, and electronics reliability. A special focus is power electronics (materials, processes, design and simulation) which is important in the electrification of the future. 

Longer lifetime and improved circularity in electronics are issues in many industries, as electronic components are everywhere. In this Blog we discuss the research area and illustrate how new scientific findings can be used. Here is a quick overview of our last findings, patents and publications in the field of predicitive maintenance, failure analysis and various methods, primarily in the field of power electronics.

Trends

Electronics is everywhere! It is what drives functionality in our modern world, and it is very difficult if not impossible to identify a single part of our society that does not in any way rely on electronics hardware in some way. It drives our manufacturing industries, enables the internet and data storage clouds, keeps our homes running with cooking, refrigeration, laundry, etc., powers our healthcare, lights our cities, keeps our traffic systems safe and enables electrical transportation. In short, we are very dependent on electronics hardware, sometimes without thinking about it.

Challenges

Electronics hardware is often very complex, it consists of many different materials (semiconductors, metals, polymers, ceramics, …) that are closely integrated in very small volumes.  

As the trend is more electronics hardware everywhere one of the challenges becomes to increase sustainability to minimize the effects on our environment. We propose solutions to this challenge.

For places where the electronics are not subjected to harsh environments we are working on recyclable materials. The target is to find sustainable materials and processes so that as the lifetime of the electronics is used up or it has completed its purpose, we can simply recycle it with the other paper recycling.

If you cannot use recyclable electronics hardware the best ways to reach a more sustainable product is to:

Make it last – long lifetime reduces the need for using materials and energy to manufacture replacement systems.

Make it modular – so that it can easily be replaced when it fails or when it needs an upgrade 

Make it predictable – understanding the remaining useful life of the electronics leads to more efficient predictive maintenance management and opens the possibility of a safer second life of the electronics with reuse, replacement, refurbishment options.

At RISE we work on all of these paths to making electronics hardware more sustainable. For example, we have developed models based on AI and physics-models to predict failure in power electronic devices during power cycling. Within the first thousands of power cycles the models can predict how many cycles remain until failure several hundreds of thousands or millions of cycles away. Not having to test until failure, but rather test until the initial onset of failure has the potential to safe time and energy in the development and qualification stages of the electronics hardware.

What we really would like is to predict the lifetime of the devices in for example an inverter used in an electric vehicle. For this purpose, we have recently developed a new, patented method to link tests in the lab to lifetime assessments in products in the application.

In short, the method includes degrading devices/components/systems to different known states of remaining life then integrating them into the target product and performing realistic use tests on the product level. The test results then serve as a database, or a “fingerprint” of how a product with degraded devices/components/systems responds in realistic use situations. Determining the remaining useful life of the devices/components/systems in the products in the field simply becomes the task of comparing their “fingerprints” to that of the database.         

Presentations, patents and publications

File 1 Research paper: A digital twin design to empower the (maritime) port as a decisionmaker in multi-organizational settings to proactively plan and optimize its utilization of resources

File 2 This is a presentation from Aug 2024 of the offering from RISE in the field of electronics, with focus on circularity and energy consumption, and how the lifetime of electronics can be improved and monitored with advanced methods (predictive maintenance and AI). We are experts on how to build electronics and how electronics fails. Focus areas are wide-bandgap power electronics and prognostication of failure/determination of remaining useful life.

File 3 Research paper with novel approach to improving prognostics model development for prediction (better maintenance and lifetime for power devices)- an invention in the field of degradation based on junction temperatures.

File 4 Research paper presented at Proceedings of the 33rd European Safety and Reliability Conference in 2023 - In this work, a data driven approach to the prognostics problem is presented. Machine learning models are trained to predict the RUL (remaining useful life) of wire-bonded silicon carbide (SiC) (MOSFETs) subjected to power cycling until failure. Three neural network architectures were trained and evaluated.