Korte beskrivelser av pågående ph.d.-prosjekter:

Measurement Strategies for battery-less, smart sensing instrumentation

Measurement Strategies for battery-less, smart sensing instrumentation

PhD candidate: Sondre Ninive AndersenSondre Andersen

Main Supervisor: Geir Mathisen

One way of reducing the cost of infrastructure maintenance is to continuously monitor the health of the structure using sensor networks. This approach allows remote monitoring, reducing the need for periodic inspections, lowering cost. However, typical sensor networks with sensors connected through wires to a data acquisition unit are expensive, both in equipment cost and in installation. New technologies have made it possible to employ wireless sensor networks, with inexpensive sensor nodes spread across the structure, all communicating wirelessly to a base station. This reduces the cost of installation, allowing such cost-saving networks to be employed more widely than before.

However, while data can easily be transmitted from wireless sensor nodes, supplying the nodes with energy is more challenging. Batteries can be used, of course, but these would eventually run out, and so would need to be replaced. Instead, ambient energy can be harvested from the environment, either from light, wind, temperature differences, vibration, or other sources. While this is possible, the amount of energy that can be extracted is usually very limited, so it is critical that the sensor node is as efficient in its energy consumption as possible.

The goal of my project is to formulate a general method for modelling the factors that impact the energy use of such a sensor node, and work towards a solution to this optimization problem.

Hydraulic effect of partially blocked and tapered inlets for circular culverts

Hydraulic effect of partially blocked and tapered inlets for circular culverts

PhD candidate: Joakim SellevoldJoakim Sellevold

Main supervisor: Nils Rüther

In a changing climate, it is increasingly important to protect infrastructure against floods. One of the most commonly used flood protection measures is the culvert: an open conduit that allows water to flow safely through embankments.

In recent years, stricter design criteria for flood protection of roads have been implemented in Norway. A large percentage of existing culverts is therefore likely under-designed, meaning they will not be able to safely handle larger floods. The present project will expand existing design methods, by investigating the effects of partial blockage and tapered inlets on flow of water through culverts. 

Main goal 1 – Effects of partially blocked culverts:

It has been documented that blockage of culverts reduces discharge capacity, causing increased local flooding and erosion in the area near the culvert. However, the present design methods do not account for the effect of partial blockage of the inlet in a precise manner. The first goal of the present project is therefore to adapt the current design methods to account for the effects of various blockage scenarios. This allows for better risk analysis related to blockage of culverts during floods.

Main goal 2 – Optimized tapered culvert inlets:

Tapered inlets have a funnel-shaped transition from the inlet section to the main barrel of the culvert. The use of tapered inlets can increase discharge capacity significantly for culverts under certain flow conditions. Such designs have been used at least since the 1960s, leading to increased performance and reduced costs. However, existing tapered inlets are of simplified designs, adapted to older production methods. The second goal of the present study is therefore to develop novel, standardized tapered inlets that can be integrated into existing product lines. The designs will be optimized for hydraulic performance. This will increase the discharge capacity without increasing the overall pipe diameter of the culvert, leading to reduced costs and carbon footprints. 

Research methodology and implementation of results:

For both partial blockage, and tapered inlets, physical scale models and CFD-simulations will be used to study the flow of water for different blockage scenarios and inlet designs. From the results, design values will be obtained, for use with existing design frameworks. This approach allows for rapid implementation of the results in existing guidelines and design manuals.

Interpreting geotechnical soil investigations using machine learning

Interpreting geotechnical soil investigations using machine learning

PhD candidate: Sigurdur Már ValssonSigurdur Mar Valsson

Main supervisor: Gudmund Reidar Eiksund

Geotechnics describes the behavior of soils and rocks from an engineering perspective. Knowledge about the grounds mechanical properties forms the basis for all geotechnical design. Soil conditions in each project are explored with soil investigations, which include borings with geotechnical drill rigs as well as laboratory tests on soil samples.

In 2013 the Geological Survey of Norway launched a national database for soil investigation (known as NADAG), where both the public and private sector could share data from their projects. The initiative was a great success, and in early 2019 the database contained registrations from over 100.000 boreholes in Norway. The Norwegian Public Roads Administration has its own database (GUDB) that is synchronized with NADAG each night. All data from NADAG (including data from GUDB) are freely available for download online.

Machine learning is a scientific field with the aim of giving computers the ability to learn and act based on data, with minimal human intervention. The aim of this project is to utilize machine learning to study a large amount of previous soil investigation data to create models that can be used for automatic interpretation of new investigations, by using both simple (decision trees, k nearest neighbors, etc.) and advanced mathematical models (different types of neural networks).

The Norwegian Public Roads Administration has conducted a pilot project to investigate the feasibility of using machine learning in geotechnics. We found that a simple machine learning model outperformed conventional methods in identifying the presence of quick clay with data from the cone penetration test (a common soil investigation method).

The main goal of this project is to study how machine learning can be used to interpret soil investigation data, as a human would approach the problem. To train models to study the data structure from each test (registered values, changes with depth, sharp transitions, etc.) to assign layers, identify soil types and to assign mechanical properties.

Interdependencies in Multi-Infrastructure Asset Management

Interdependencies in Multi-Infrastructure Asset Management

PhD candidate: Shamsuddin DaulatShamsuddin Daulat

Main supervisor: Franz Tscheikner-Gratl

Urban infrastructure is ageing and in constant need for maintenance and rehabilitation. Maintenance and rehabilitation are complex and often costly. Researchers and practitioners are constantly trying to find cost-effective solutions to achieve the highest impact on the overall condition of the networks while keeping or improving the current performance and service levels. Rehabilitating multiple co-located infrastructures (e.g., roads, sewer, water, electric cables, etc.) simultaneously has shown promising results. There are potential synergies in integrating the rehabilitation of these infrastructures, which can be exploited. Exploring and evaluating these possibilities and implementing them into a decision framework is the subject of this PhD project.


This PhD project will investigate on how to integrate the rehabilitation of multiple co-located infrastructures by focusing on the following topics:

  • Identification of the infrastructures which have the potential to be rehabilitated simultaneously and mapping out possible dependencies, interdependencies and externalities.
  • Finding the optimal timeframe for this integrated rehabilitation considering the varying life expectancies of the various infrastructure.
  • Costs and benefits of the integrated rehabilitation approach and evaluation of the necessary savings due to possible synergies to make the approach cost effective.
  • Evaluation of current performance assessment methods and possible changes on the performance of individual infrastructures when multi-infrastructure rehabilitation is practiced.

Condition monitoring of drainage systems

Condition monitoring of drainage systems

PhD candidate: Clara Weberclara weber

Main supervisor: Inge hoff

Water and moisture have a significant influence on pavement performance and traffic safety. An efficient and properly working drainage system is therefore a main requirement to keep the water away from the road and its structure.

Drainage systems consist of different elements including permeable pavement layers, drains, culverts, and ditches. If one of the elements is out of order the whole system fails. The facts that a major part of the drainage system is old, the documentation is insufficient, and the current maintenance procedure is laborious and cost-intensive often leads to major incidents.

The goal of my research is to find a smart way to monitor subsurface drainage systems. By finding an efficient procedure, defects can be identified before the incident occurs. This will lead to fewer costs and safer roads.

To start with, a smart solution to detect, locate and document subsurface drainage elements such as culverts will be developed. The next step will be to find a procedure to analyze the condition of the culverts and other drainage elements. The suitability of remote sensing equipment more precisely Ground Penetrating Radar (GPR) will be tested for localization as well as condition analysis. The tested methods will be evaluated and optimized under real and prevailing local conditions.

Reassessment of structural integrity of road bridges – with special emphasis on data acquisition and analysis and maintenance planning

Reassessment of structural integrity of road bridges – with special emphasis on data acquisition and analysis and maintenance planning

PhD candidate: Frida LiljeforsFrida Liljefors

Main supervisor: Jochen Köhler

Maintenance of existing road bridges is needed to ensure that the capacity of the bridge is sufficient in relation to the loads acting on it throughout the specified service life. Given satisfactory maintenance, the existing infrastructure can continue to serve the needs for transportation in a sustainable way.

When in doubt about the structural state of a bridge, it must be reassessed. An ageing bridge is a complex system that involves a lot of uncertainties that require careful consideration. Depending on the outcome of a reassessment, different actions may be taken, such as repairing or replacing the bridge or gathering more information. As rehabilitation of bridges is expensive, there is a need for effective methods and strategies to prioritize among possible investments.

In this PhD-project, a decision-making framework for road bridge maintenance will be developed. Basis is taken in the current Norwegian, European and international best practice, along with well-established theories on reliability based and risk based structural reassessment, Bayesian decision making and value of information, among others. Implementing efficient maintenance strategies ensure safe bridges using the least possible amount of resources by doing the right action at the right moment.

Machine learning and computer vision for smart maintenance of road infrastructure

Machine learning and computer vision for smart maintenance of road infrastructure

PhD candidate: Mamoona Birkhez ShamiMamoona Birkhez Shami

Main supervisor: Frank Lindseth

Roads are a vital infrastructure of a country. They require periodic checks and maintenance to ensure unhindered travel between cities and communities which is imperative for a functioning society. Given the limited resources, the decision of when and where to delegate them is very important. Traditionally road maintenance is done by manual inspection or through specialized cars which is costly. Given the modern technological advances, a cost-efficient, scalable and intelligent solution that can predict maintenance needs can be built using state-of-the-art Machine Learning techniques.

This PhD project has three major goals:

  1. To collect data from local authorities and label the typical types of road damages. This is a labor-intensive task as each image must be labelled manually and a bounding box drawn on the image. This will generate a Dataset which will serve as Ground Truth for training purposes.
  2. To develop a data analysis tool using machine learning methods. This tool will be able to identify the extent and type of road damage present in a road segment given the images of that road segment.
  3. To develop data-driven hybrid models for predictive maintenance to provide decision support to the Road Authorities.

Causal Inference in Transportation Economics using Machine Learning Approaches

Causal Inference in Transportation Economics using Machine Learning Approaches

PhD candidate: Cong CaoCong Cao

Main supervisor: Colin Green

Motor vehicle emissions are a major source of air pollution in cities worldwide. The air pollution caused by traffic and the consequent health effects have generated increasing attention. It has long been recognized that measures such as increasing road capacity or prioritizing public transport do not themselves solve these problems, and that optimal policy responses rely on better aligning the private benefits of transportation to its social costs. A critical input into this is an understanding of the underlying relationships between transportation and these social outcomes. While a focus of much research, in practice we have an imperfect understanding of these relationships. Hence, a focus of my dissertation is to utilize approaches, drawn from recent advances in machine learning and related areas, to improve our understanding of the economic and social effects of transportation, and transportation policies.

With the advent of machine learning algorithms, the boom in artificial intelligence technology has spread across various scientific disciplines. A range of tasks previously difficult to handle are now feasible through the use of machine learning. Of particular relevance, there has been the recent development of a range of machine learning methods and approaches that aim to allow causal inference in complex large observational data settings. These methods seem particularly applicable to transportation research, yet currently there exists little research in this area. 

This project will lead to several important outcomes. These include both methodological advances, but also valuable scientific input into ongoing debates regarding road network maintenance and development, and the formulation of a policy aimed at mitigating the adverse effects of traffic.

Strategies and criteria for preventive and corrective maintenance

Strategies and criteria for preventive and corrective maintenance

PhD candidate: Tianqi SunTianqi Sun

Main supervisor: Jørn Vatn

Norway has a fairly complex road network due to the complicated geographic conditions. Billions of NOK are spent each year on its operation and maintenance, as well as renewal and upgrading of the road infrastructure. Maintenance itself contributes to almost half of these expenses. It is therefore of great value to choose effective maintenance strategies, so that the invests make the best possible contribution to a safe and effective road network.

This project focuses on the development of economic models for road maintenance from a reliability perspective. Taking the cost of various maintenance actions and the system reliability into account, these models provide a framework for the evaluation of different maintenance strategies, which provides basis for decision makers.

A set of degradation and failure models for specific road elements shall be developed to capture the characteristic of their deterioration process and allow a reasonable simulation of its propagation. With more information gathered through the inspections, the above models could be continuously updated to have a better fitting with the real case. Different maintenance strategies, for instance different inspection intervals and degree of repairs, could then be formulated mathematically to incorporate with the economic models. The relevant economic and reliability results for each strategy could be calculated and compared.