Master Thesis: Optimizing Straightness Measurement Algorithm for Bars via Image Analysis - Hiring in process/Finished, not possible to apply

What can you expect?

You will become part of a company that does the right thing simply because it’s right—a company with a market-leading CO2 position and a strong commitment to creating a sustainable society.

Ovako is an employer for those who value the efficient decision-making of a smaller organization, combined with the security of a larger one. At Ovako, you can expect a fast-paced, forward-looking workplace with an open atmosphere, a strong sense of camaraderie, and a highly driven organization. As an employee, you will be offered:

• A workplace focused on employee health, safety, and well-being
• Significant responsibility and freedom in how you approach your tasks
• A culture that embraces challenges and welcomes ideas for greater efficiency
• The opportunity to work in a dynamic environment alongside dedicated and experienced colleagues

Background and purpose

In the steel industry, crookedness of bars presents a significant hidden challenge. Currently, there is no cost-effective method for inline measurement of straightness, leading to reliance on visual inspection. This can cause issues such as customer complaints, equipment damage, production disturbances, and even reduced production speed. Straightening processes, often implemented to rectify the bars, are also a costly bottleneck.

The vision is to introduce a cost-efficient inline straightness measurement system capable of classifying bars into three categories:

a) Sufficiently straight: Directly sent to the customer without further processing.

b) Slightly crooked: Requires straightening operations.

c) Too crooked: Discarded as scrap.

Besides sorting, straightness measurement provides valuable feedback to the process responsible for the crookedness, allowing for improvements over time and an increase in the percentage of sufficiently straight bars.

Research question and purpose

The goal is to optimize the existing algorithm for straightness measurement, achieving an accuracy of approximately 0.2-0.3 mm/m for bars up to 10 meters in length. The current version of the algorithm is based on a modified U-Net convolutional neural network architecture. The algorithm processes high-resolution images of flat steel bars, segmenting each pixel to classify it as part of the bar or the background. Once this segmentation is completed, the straightness measurement is performed.

The focus of this thesis is to optimize the algorithm to ensure faster calculations without compromising the quality of the measurement, as the system is intended for real-time "in-line" measurement, where time is a critical factor.

Research question: How can the algorithm be optimized to improve processing speed? What adjustments can be made without sacrificing the accuracy of the straightness measurement? How does the optimized algorithm perform under the constraints of real-time processing?

Supervision and Collaboration: This thesis will be conducted in close collaboration with the production and quality departments, with support from IT department. Regular meetings will be held to ensure alignment with the project's goals and timelines.

Expected Outcome: An optimized algorithm that processes straightness measurements faster without losing accuracy, enabling its use in a real-time production environment.

Your Profile

The ideal candidate should possess knowledge and experience in deep learning and Python, preferably with skills in image analysis. The student should be able to work independently while collaborating with the production and IT departments to implement and test the optimized algorithm.

Information & other

Compensation: Ovako will compensate you for completed master thesis corrsponding to a value per "högskolepoäng".  We can also in most cases assist with accomodation during the time you are performing your master thesis with us. Ovako will support with travel to and from our sites for interviews, eventual trips back to the university for meetings with supervisor and equivalent.An optimized.

Recruitment: Selection will be ongoing and during the process you will get the chance to meet potential supervisors, relevant managers and stakeholder. The process will include ability tests, interviews, background checks and references.