Research team
Predictive Maintenance for Non-Road Mobile Machinery
Abstract
Modern non-road mobile machinery (NRMM), such as equipment used in agriculture, earthmoving, lifting and hoisting, construction, and container and bulk handling, is equipped with numerous sensors as standard. The sensor data required for the proper operation of these machines is typically communicated to their various subsystems via a CAN network. Since the introduction of the J1939 standard, the communication layer of the CAN protocol used in trucks and NRMM applications has been partially standardised. This makes it easier to retrieve sensor data from the network and use it, for example, to generate additional operational and strategic insights. Capturing and purposefully using all these sensor data represents an enormous opportunity for NRMM operators, but this potential remains underutilised today. Furthermore, prior interaction with the target group has shown that unplanned machine downtime is a critical cost factor that directly puts pressure on profit margins in the sector. From an economic perspective, any tool that can reduce or prevent such downtime is therefore essential. Predictive maintenance (PdM) offers a solution by using advanced data analysis, including artificial intelligence and machine learning, to predict failures before they result in costly interruptions to operations. Although the academic literature has extensively demonstrated the effectiveness of sensor-data-based PdM, its adoption in Flemish industry remains limited. The main barrier to adoption is not a lack of domain knowledge. Technical staff at the target companies are highly aware of the most common defects and their associated financial impact. However, in-depth knowledge, practical skills, and awareness of the enormous potential of AI for using J1939 data to proactively prevent these known issues remain insufficiently developed. There is an acute shortage of professionals capable of bridging the gap between in-depth machine knowledge and advanced data analysis. This is partly because NRMM operators are not yet actively seeking such profiles on the labour market. As long as this synergy is lacking among technical staff and operators, the technological opportunity will remain unrecognised, and demand for external expertise or internal innovation initiatives will fail to emerge. This project aims to bridge precisely this gap by translating the use of J1939 data and advanced AI capabilities into the familiar working practices of technical personnel. In this project, Karel de Grote University of Applied Sciences and Arts builds on the expertise recently developed within its Centre of Expertise for Sustainable Industry in decoding and analysing CAN and J1939 data. This expertise was acquired through the RevCAN, U-CANsim, and CremAN projects, which successfully achieved brand-independent data capture and the simulation of vehicle and machine data for industrial applications. The Industrial Vision Lab (InViLab) at the University of Antwerp builds on recent research in probabilistic machine learning and data-driven modelling. It will apply this expertise to translate J1939 sensor data from NRMM into robust predictive-maintenance applications. In summary, the large volume of available sensor data offers a unique opportunity to further reduce NRMM downtime through predictive maintenance. However, fleet operators and maintenance providers currently lack sufficient awareness and expertise to effectively unlock and capitalise on this technological potential.Researcher(s)
- Promoter: De Boi Ivan
- Co-promoter: Ceulemans David
- Co-promoter: Vanlanduit Steve
Research team(s)
Funding
- VLAIO
Project website
Project type(s)
- Research Project
Wireless capsule endoscopy based on Gaussian process latent variable models.
Abstract
Endoscopy plays a pivotal role in both diagnostic examinations and minimally invasive surgical procedures. A special type is wireless capsule endoscopy, where patients ingest a small pill-shaped camera. Despite its importance, the endoscopic images and videos exhibit serious drawbacks, such as substantial distortion, low resolution, missing frames, specular reflections, and so forth. In this project, I will tackle several of these challenges. In order to do so, I will first develop a novel endoscopic camera calibration procedure. Next, based on this, I will adopt approaches from the field of Gaussian process latent variable models and the world of generative AI in general to formulate models that construct an alternative latent space representation of the data. Probabilistic machine learning models, such as Gaussian processes, offer interpretability (no black-box), which is especially crucial in evidence-based medicine as it offers transparency and helps build trust with clinicians. Improved camera calibration and innovative perspectives on latent spaces hold the potential to revolutionise various techniques, including 3D trajectory estimation, mosaicking and many more. As a result, my research stands to significantly enhance the precision and efficiency of clinicians when interpreting endoscopic images. This, in turn, promises to elevate detection rates, enhance the accuracy of abnormality size measurements, and contribute to the advancement of minimally invasive surgery.Researcher(s)
- Promoter: Penne Rudi
- Fellow: De Boi Ivan
Research team(s)
Funding
- FWO
Project type(s)
- Research Project
Implicit camera system calibration through Gaussian processes.
Abstract
Camera calibration is a foundational step in accurate 3D measurements. However, direct calibration can be complex and time-consuming, especially for multi-camera systems with varying configurations. This research proposes an implicit calibration approach using Gaussian processes, a probabilistic machine learning technique. By learning the mapping between 2D image pixels and 3D world coordinates, we can bypass the need for explicit calibration of intrinsic and extrinsic camera parameters. The proposed method is particularly advantageous for challenging scenarios, such as wide-angle lens systems and multi-camera setups with diverse camera types (e.g., RGB and infrared). By leveraging Gaussian processes, we can handle low-data regimes and provide reliable uncertainty quantification. We aim to demonstrate the effectiveness of this approach through two use cases: 3D 180°-360° Environment Mapping: Using multiple wide-angle RGB cameras to capture a comprehensive 3D representation of a scene. 3D Thermography: Combining RGB and infrared cameras to generate 3D thermal maps, enabling detailed analysis of heat distribution. This research will contribute to advancing the state-of-the-art in 3D computer vision by providing a flexible and robust solution for implicit camera calibration, enabling accurate and efficient 3D measurements in various applications.Researcher(s)
- Promoter: De Boi Ivan
Research team(s)
Funding
- BOF
Project type(s)
- Research Project
A general line variety model for sensors, allowing stable calibrations that meet the accuracy standards for medical applications.
Abstract
The popular pinhole model for imaging sensors and the associated calibration procedures appear to be inadequate for some of the new generation sensor technology. Even for classical RGB cameras, this standard model leads to unstable calibrations, with the need for an extra model to remove lens distortion. We propose line varieties as a unifying modelling for a broad set of sensors. As opposed to other previously published attempts in this direction, we identify the sub-varieties that correspond to real sensors. This enables us to extend interpolation techniques and Gaussian processes, to support sensor calibration from small samples of lines. We aim fundamental contributions to the fields of Line Geometry and Probabilistic Numerics. Our goal is to develop the framework for multi-sensor configurations (laser scanners, IR-cameras,…), providing measurement fusion, using the developed line models, and to achieve accuracy levels for sensor-supported Radiotherapy.Researcher(s)
- Promoter: Penne Rudi
- Fellow: De Boi Ivan
Research team(s)
Funding
- BOF
Project type(s)
- Research Project