Camilla Munther

Let’s start with your story: What’s your background, and how did you end up working on this particular research topic?
I have a background in automation and production engineering. My first real contact with industrial maintenance came about 15 years ago. During my bachelor’s degree in automation and mechatronics, I worked part time as an automation engineer installing patented technology for sootblowing. The system didn’t just change how the sootblowers operated — it also provided valuable data to the maintenance department. I remember a technician’s reaction when we showed the new interface: 'You mean we’ll get an alarm if the steam valve doesn’t open as expected?'. Today, many of us take that kind of alarm for granted. For them, at that time, having access to real-time data on expected versus actual steam flow was revolutionary. It meant they could act proactively, preventing serious damage to expensive equipment. So, even if the main selling point for this new technology was for operational purposes, it became clear that it also could be used to increase the maintenance performance. But that required a change of their work processes.
Maintenance wasn’t a major topic during my university studies, but I ended up doing my master’s thesis on quantifying the effects of maintenance using discrete event simulation. A few years later, I got the opportunity to start a PhD at Chalmers University of Technology, working in a research group focused on production service and maintenance systems. My PhD journey was driven by a desire to understand how people, processes, and strategies must align to successfully implement new technologies and ways of working in maintenance. This led me to explore Smart Maintenance as an organisational innovation — a perspective that I believe is essential for meaningful and sustainable change.

In a nutshell, what’s your thesis about — explained as if to a maintenance professional over coffee?
My thesis is about helping maintenance organizations and all targeted employees become more skillful, consistent, and committed to working with Smart Maintenance. The key is to treat Smart Maintenance not only as a technical upgrade, but as an organisational innovation.
By viewing Smart Maintenance through the lens of innovation theory, we can better understand how to implement it. In my thesis, I use five innovation characteristics: relative advantage, compatibility, complexity, trialability, and observability. Financial calculations alone aren’t enough — there must be a true belief in the relative advantage of Smart Maintenance. Compatibility means aligning initiatives with existing values and norms, starting at a point that fits the current state of the organization.
Complexity can be reduced by breaking down change into smaller, manageable steps, which also increase the trialability that allows experimentation. Observability ensures that progress and results are visible and measurable. The maintenance manager’s task and responsibility become to lead people in change, rather than being a technical leader.
In my thesis, I propose a cyclical, six-step process that supports Smart Maintenance implementation:
1. Benchmark the organisation.
2. Set clear goals.
3. Define strategic priorities.
4. Plan key activities.
5. Elevate implementation.
6. Follow up.
Combined with the insights gained from the perspective of organizational innovation, this process can be used as a framework to guide organizations in being more skillful, consistent, and committed to Smart Maintenance.

From the lab to the shop floor: How could your findings be applied in real-world maintenance or asset management settings?

In Sweden, we benefit from strong collaboration between industry and academia, which allows research to be conducted very close to the shop floor. I’ve worked closely with several industrial companies, and the cyclical process I propose in the thesis is designed to be applicable by industrial maintenance managers. Smart Maintenance implementation will look different in every organisation, but my findings offer a framework to follow.

A handful of industry practitioners have literally read each word in my thesis. From cover to cover. As a researcher who is driven by industrial impact, that is probably one of the most awarding compliments you can get from industry. I see this as a validation that my research is relevant and applicable for real-world settings.

Research is rarely a straight road: What were the toughest hurdles you faced, and how did you overcome them?

To be honest and a bit personal: the vulnerability that comes with doing a PhD is tough — the feeling of constantly exposing your thinking and work. I became very aware of the importance of how I expressed myself (both in text and speech), always striving for clarity and quality. It’s a demanding process, both mentally and emotionally. Luckily, I was surrounded by amazing people. My supervisor and colleagues I had during the PhD studies were always encouraging and supportive. Their feedback helped me refine my ideas and continue to always try to do a bit better.

Industry 5.0 is all about people, sustainability, and resilience: Where do you see your research fitting into this bigger vision?

My research fits naturally into the Industry 5.0 vision because it emphasizes the human side of technological change. Smart Maintenance, when treated as an organizational innovation, becomes a way to empower people — not replace them. It´s about making the whole organization more skillful, consistent, and committed to Smart Maintenance. It supports resilience by helping organizations and all targeted employees adapt to change, and it contributes to sustainability by enabling more efficient and proactive use of resources.

Inspired by principles from innovation management, my work encourages maintenance leaders to foster creativity, challenge existing routines, and actively seek untapped value. It’s about building a culture of learning and exploration. This aligns with Industry 5.0’s emphasis on human-centric, sustainable, and resilient production systems.

Women in maintenance and asset management: From your perspective, how can the sector attract more women and create an environment where they can thrive?

We need to highlight female role models, offer mentorship opportunities, and foster inclusive cultures where different perspectives are valued. It’s also important to challenge outdated stereotypes and show that maintenance is a dynamic, forward-looking field where women can lead, innovate, and make a real impact. Maintenance and asset management are no longer just about fixing machines — they’re about strategy, innovation, and people. This appeal to a diverse range of professionals, including women, who bring valuable perspectives to the field.

What significance has receiving the EFNMS Thesis Award had for you personally and professionally?

Receiving the EFNMS Thesis Award was a great honor! Anyone who has done a PhD — or supported someone through one — knows the level of commitment it requires. To have that work recognized at a European level is incredibly validating. Personally, I´m full of pride and motivation. Professionally, it led to a wider network and potential collaborations. I look forward to advancing maintenance and asset management together with my new contacts!

Looking ahead: If you could choose, what would be the next big challenge for researchers and industry professionals to tackle in your field?

Rather than a single “next” challenge, I see a continuous and evolving one: integrating Smart Maintenance into broader organizational strategies. We need to move from asking “How do we implement this technology?” to “How do we evolve our organization to unlock its full potential?” That means developing new models, metrics, and mindsets that reflect maintenance’s expanding role in digitalized and human-centric production systems. Inspired by innovation management, maintenance leaders should be encouraged to explore untapped value rather than solving the problems we have today, foster cross-functional collaboration, and formulate a clear vision for how maintenance contributes to the company’s future. This shift requires time, resources, and a willingness to challenge traditional ways of thinking — but it’s essential for long-term competitiveness and sustainability. Like innovation leaders promote freedom to explore and experiment, maintenance leaders must create space for discovering untapped values — and connect maintenance to the company’s broader vision and competitiveness.

Your own next chapter: What’s next for you — more research, industry work, teaching, or something else entirely?

Right now, my focus is on being a mom of two — I’m on parental leave and enjoy this chapter of life. At the same time, I’m staying connected to the field, keeping an eye on ongoing applications for new research projects, as well as planning a conference presentation. From January 2026, I’m excited to return to research and I look forward to collaborating with other researchers in the field, as well as industry partners, aiming to continue contributing to the development of maintenance as a strategic and innovative function.

Aleksanteri Hämäläinen

Let’s start with your story: What’s your background, and how did you end up working on this research topic?

I first got into coding in high school at the Päivölä School of Mathematics, which led me to study computer science at Aalto University. I was similarly first introduced to AI and deep learning in high school, and it has fascinated me ever since. That’s why I ended up majoring in Machine Learning, Data Science and Artificial Intelligence.

As I was searching for a topic for a thesis during the last year of my master's, I was contacted about a topic on AI and condition monitoring by a doctoral researcher I had worked with earlier on a group project. I had pretty much no experience in mechanical engineering but was promised that the topic was very interesting on the AI side of things and that the data was good. In retrospect, I very much disagree with the quality of the data, but realizations about the problems with the data have become the next interesting topic I have pursued, so it’s not like I’m complaining. As a side effect, I’ve also ended up learning quite a bit about mechanical engineering and signal analysis, which I’m certain will be useful in the future.

In a nutshell, what’s your thesis about — explained as if to a maintenance professional over coffee?

My thesis addresses the challenge of using deep learning models for condition monitoring of rotating machines, particularly when data is limited. In most cases it’s not possible to have fault data from every operating condition, such as varying rotating speeds.

Furthermore, even machines of the same model can differ because of manufacturing tolerances, installations, and usage histories. In my research, I demonstrated how few-shot learning, prototypical networks, and careful consideration of operating conditions can be used to get condition monitoring models to work well for gear fault diagnosis in scenarios not covered during the training. The findings specifically showcase good generalisation over rotating speed, which commonly varies in rotating machines, and sensor locations, which represent differences between machines. 

From the lab to the shop floor: How could your findings be applied in real-world maintenance or asset management settings?

Many companies involved in condition monitoring are already using AI in some manner or are experimenting with ways to do so. My thesis offers a way to approach some of the key challenges, particularly generalisation over operating conditions and machines. Additionally, the important parts are not overly complex to implement. Instead of encouraging others to exactly replicate what I have, I hope they will integrate my findings into their own work and ideas. I’d be very glad if my research helped someone to overcome a long-standing problem in their systems or sparked an idea of “Aha, this is how I’ll get it to work!"

Research is rarely a straight road: What were the toughest hurdles you faced, and how did you overcome them?

The ever-present problem in using deep learning for condition monitoring is the lack of high quality, publicly available datasets. The current ones generally have a good selection of rotating speeds and loads, but the ones most often used in research all lack essential elements, such as long enough samples, multiple instances of the included fault types, repeated setups, or sufficient healthy data.

The results of training a deep learning model on 10 seconds of fault data and testing it on the next 10 seconds of the same run do not significantly correspond with real world performance.  The aim is not to recognise one exact fault instance, but all faults of the same type. The vibrations of a test rig in a lab do not significantly change within minutes either, so the model could be based on its predictions on the vibration signatures of the installation, manufacturing errors of a component, or even background noise.

This same problem was a concern in my thesis, and ultimately it was only partially overcome. My thesis includes results where model training and testing were conducted with sensors located in different parts of the powertrain in addition to just splitting the data by time. These changes in sensor location introduce significant changes to vibration signatures, simulating changes between two different machines. I was happy with this solution for my thesis but have since then strived for even more realistic test scenarios, by using entirely separate datasets for training and testing.

Industry 5.0 is all about people, sustainability, and resilience: Where do you see your research fitting into this bigger vision?
Wind farms are a perfect example of a setting where condition monitoring is essential for numerous nearly identical rotating machines. Increasing the uptime of wind turbines and decreasing their maintenance costs could help increase their portion of energy production.

What significance has receiving the EFNMS Thesis Award had for you personally and professionally?

I’m very honored to have received the EFNMS award. It showed that there is real interest in the topic and that what I was working on was worth pursuing further. I sincerely hope my next findings will garner similar, and hopefully even greater, interest.

Looking ahead: If you could choose, what would be the next big challenge for researchers and industry professionals to tackle in your field?

I would like to see the creation of better public datasets for research. The ARotor Lab recently published the Aalto Gear Fault Dataset, which includes measurements from multiple healthy and faulty gears and repeated installations of the faulty gears. I hope other research groups will include these elements in future datasets they publish.

However, it would be even more beneficial to have a company publish a dataset containing real fleet data. I of course understand that a lot of data may not be publishable, but I’m sure there is some data that would be at its most valuable when many researchers are working on developing new methods for it. It wouldn’t even have to be highly curated, I’m sure some desperate doctoral researchers, such as I (wink wink), would quickly make a cleaned version.

Afterwards, I would like to see improvements in testing methods to better reflect real-world usage.

What is the point of publishing papers with accuracy close to 99.99% if the results are only relevant to academia?

Your own next chapter: What’s next for you — more research, industry work, teaching, or something else entirely?

I’m currently working on a PhD at the Aalto ARotor Lab focusing on the same topic as my master's thesis, so you could say my next chapter was not very different from the previous one. As I am writing this, I am at the University of New South Wales in Australia for an exchange with a research group here, so the topic of condition monitoring has taken me to interesting places. After my PhD, I’m leaning towards industry work as the most likely next step.

As I mentioned earlier, I strongly believe that significant progress in AI based condition monitoring is best achieved with larger amounts of higher quality data, which is unfortunately limited in public research. I also enjoy working on problems where I can see more immediate real-world benefits than is common in academia. However, I am not saying that is the only option. I got into this topic by taking a chance on an interesting opportunity and that is the plan for my next step too.