Improving Magnetic System Simulation

Master Thesis
HIGH TECH CAMPUS VILLACH, EURO­PA­STRASSE 12, 9524 VILLACH

In the network of science and industry, Silicon Austria Labs (SAL) offers re­search in four pionee­ring divi­sions: Sensor Systems, RF Systems, Power Elec­tro­nics and System Inte­gra­tion. SAL – a great place to re­search. 

Background – Magnetic System Simulation

Magnetic sensor systems are used to measure observables of interest through the magnetic field. Such systems include a coil or a permanent magnet to create a magnetic field, that is then modulated by the environment. The system is completed by a sensor or sensing principle to detect this modulation.

Designing magnetic sensor systems is mainly done by simulation and computer experiments that solve the underlying field equations and attempt to map the physical properties. One major problem when designing realistic physical systems is the large number of input variables that describe the geometry and material parameters together with their tolerances. In magnetic sensor system, the computational system optimization in such large parameter spaces is a challenging topic.

In the special case of position and orientation detection systems it was shown that analytical formulas can be used with more than 99% precision. The extremely low computation times of analytical formulas make optimization in higher dimensional spaces possible. Many permanent magnet solutions were implemented in the magpylib Python library together with geometric manipulation routines for easy practical use.

Thesis

Within the course of this thesis we want to study different optimization routines for global optimization of magnetic systems with many (10-100) variables. To this end black box algorithms like genetic optimization, differential evolution, particle swarm and basin hopping should be studied, tested and compared to gradient based algorithms. The student should take over the tasks of study and compare the structure of the cost function for different magnetic and standard test problems.

In order to select the best appropriate optimization algorithm, the student should reach a clear understanding of the structure of the magnetic problems.

In addition, advantages from multithreading and powerful implementations from numpy based on SIMD for efficient computation of vectorized code should be considered to reduce the computation time of the optimization routine.

In summary, we offer to teach about magnetism and magnetic sensor systems, computation and simulation in Python, potential scientific publication and expect willingness to learn and apply this knowledge.

About us

Silicon Austria Labs (SAL) is on the way to become a top Euro­pean re­search center for elec­tronic based systems. In the network of science and industry, we carry out re­search at the highest global re­search level and thus create the basis for new types of prod­ucts and processes.  SAL is divided into four divi­sions: Sensor Systems, RF Systems, Power Elec­tro­nics and System Inte­gra­tion, and is equipped by state-of the art equipment, consisting of comprehensive optical labs, C5/C8 cleanroom facilities, backend processes for micro-packaging, as well as custom prototyping facilities (custom electronic development, 3d printers, CNC-machinery, Inkjet printer).

 

Start Date / Duration / Contract

Start date (planned): as soon as possible

Duration (planned): 6 Months

Payment: 1385€ gross (38.5 Hours/Week)

Place: Villach, Austria.

 

Profile / requirements

  • Background in computation, math, physics or similar
  • Ideally experience with scientific Python
  • Interest to learn about magnetism and global optimization problems
  • Ability to work independently AND in a team
  • High level of english

 

Contact

Michael.ortner@silicon-austria.com