É«É«À²

The two test devices have been built to help study the frictional, wear, and thermal behaviors of marine thruster lip seals.  Due to their relatively large size (designed to run on shafts 300 mm), these seals are seldom studied in the literature. The test devices offer a robust means to investigate the behavior of marine thruster lip seals, particularly their difficult to estimate thermal behavior.

Movement of piston rings in relation to piston in an internal combustion engine affects the transport of lubricating oil into the combustion chamber. Lubricating oil interferes with the combustion event and causes issues when using greener fuels such as hydrogen. Piston ring dynamics are investigated in a static laboratory scale testbench that provides a more controlled environment for developing measurement methods for piston ring position.

Vibrations can be harmful for critical components in rotating machinery. Wire rope isolators can be used to keep vibrations at acceptable levels. They have good damping properties due to frictional losses between wires, and their nonlinear stiffness properties are beneficial in isolation applications. This research aims to better understand how the beahvior of wire rope isolators is affected when they have been in use for a long time. Possible changes in, for example, effective stiffness are critical to be aware of when designing wire rope isolators for industrial applications. So far wire rope isolators have shown excellent capability of attenuation attenuating lateral vibrations. Yet, they have not been investigated thoroughly in torsional applications. Therefore, this research also aims to explore whether wire ropes could be used to effectively attenuate torsional vibrations.

Eddy current sensors measure relative distance to the shaft, determining shaft movement. However, the measurement is affected by electrical runout. Electrical runout is a measurement error in eddy current sensors, caused by variations in the material properties of the target. Electrical runout is a significant error cause when measurements are taken with micrometer accuracy. The research focuses on how this electrical runout could be reduced with diamond burnishing.

Air bearings are gas lubricated bearings, which enable high-speed precision motion with low friction. Air bearings have interesting new applications in production machinery, where increasing demands on quality and energy consumption make traditional solutions infeasible. The research at ARotor has improved the understanding of the manufacturing process and performance of air bearings. Current research topics include the use of porous graphite aerostatic bearings as a sensor and manufacturing of gas bearings with metal 3D printing.

The quality of paper can be analyzed by measuring paper samples in the laboratory or on the paper machine during production. In our laboratory, we use a paper analyzer that allows us to measure the thickness of paper from samples that are kilometers long. Thanks to the measuring device, we obtain data for analysis that allows us to assess the performance of the paper machine. Therefore, paper quality reveals important information about the machine's operation, such as the condition of the rolls, vibration, or adjustment issues.

AI can be used to monitor rotating machinery for both diagnosis (current condition) and prognosis (failure prediction) based on sensor data. Training such models requires data from both healthy and faulty conditions. This is generated using a test bench—a scaled maritime thruster—where realistic loads are applied via a load motor, from steady operation to simulated ice impacts. The dataset includes multiple gear faults and is collected using sensors such as accelerometers, torque transducers, and encoders.

The aim of virtual sensing in maritime propulsion systems is to estimate quantities, such as torque and rotational speed, which are conventionally measured using physical sensors. Virtual sensors provide information from locations on the drivetrain that are inaccessible with physical sensors. 

The aim of the project is the complete automation of the pedestrian labelling process. The process leverages the stereo alignment of the images to: 1) Detect pedestrians in infrared 2) transfer the detection seamlessly to RGB since the frames are aligned 3) return a labelled dataset for pedestrian detection.

Dimitrios Bouzoulas

Description: This project focuses on improving material handling in factory intralogistics through three key ideas. First, a tracking system combines camera and UWB data to locate objects accurately. Second, a vision-based quality inspection system uses synthetic data to detect defects with limited real data. Third, an autonomous mobile robot is developed to detect a trolley, navigate to it, and attach for automated transport. Together, these solutions show how sensing, perception, and robotics can improve efficiency and automation in industrial environments.

Lucas Foley, Aadesh Chaudhari, Muhammad Anis

Simultaneous localization and mapping (SLAM) allows a robot to navigate in a previously unknown environment, making a map as it goes. A SLAM method only using one camera as input would be very cheap, but a single source of images cannot know the scale of whatever is in the images or how far they are without any other information. In this research, I experiment with adding machine learning –based depth estimation to help visual SLAM localize in traditionally hard environments.

Eelis Peltola

This project studies how an overhead 3D LiDAR sensor can be used to detect people below it in an indoor crane workspace. We use AI-based 3D perception methods to identify and then track human targets who are under a predetermined vicinity from the sensor. The results show promising performance and execution speeds, to contribute towards safer human-aware automation in similar industrial workspaces.

Nilusha Jayawickrama, Henrik Toikka, Risto Ojala

In this project, we explore how future automated vehicles can work together (V2X technology) to detect road users even when they are somewhat hidden or far away, especially in urban driving scenarios. Instead of relying on just one car’s sensors, we combine information from multiple vehicles in an aim to contribute towards a smarter and safer system. To make this possible, we are building Scale Down. It is a unique platform that is both in a digital as well as physical formats to enable safe testing of challenging driving situations that are difficult or dangerous to study in real life.

Nilusha Jayawickrama, Aleksi Pippuri, Jie Lie, Risto Ojala

This work presents a few-shot object detection approach that uses vision foundation models to detect unseen industrial objects from only a few samples. It eliminates the need for retraining by separating object localization and identification through prototype-based matching.

Hari Prasanth S.M.

This work presents a method for creating 3D maps of CERN facilities for autonomous radiation surveys using a non-repetitive LiDAR. Compared with repetitive LiDAR, non-repetitive LiDAR can produce denser and more information-rich representations of the environment. However, it also makes registration and mapping less accurate. This study addresses the inaccuracies associated with non-repetitive mapping methods.

Pejman Habibiroudkenar

Method that can automatically label the road in challenging winter conditions just based on the driving history: areas which have been driven before are used to teach a machine learning model which labels similar parts of the world as drivable. Increased robustness is achieved by fusing lidar and camera into a 3D map. The work is important for enabling automated driving in winter conditions.

Eerik Alamikkotervo

Our powertrain emulator is a hardware platform used for experimental research in electric mobility. It aims to replicate the behavior of systems between the battery and electric motors. In doing this, we can study vehicle behavior in different driving scenarios in a safe and controlled environment.

Kaarlo Mäkelä, Jenni Pippuri-Mäkeläinen, Jesse Pirhonen, Haoyu Song, Risto Ojala