kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Machine learning-based adjustments of thermal networks
KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Chemical Engineering.
2022 (English)In: 11th International Conference on Power Electronics, Machines and Drives (PEMD 2022), 2022Conference paper, Published paper (Refereed)
Abstract [en]

A-priori defined thermal networks usually show reasonable agreement between measured and simulated temperature results. Discrepancies are, however, encountered on a regular basis depending on how well the model is replicating reality, and as a result, the accuracy of the thermal network may be reduced. This paper shows an evaluation of machine learning methods for adjusting the thermal network of an electric machine based on temperature inputs. The thermal resistances constituting the network are thus updated with regards to the operating conditions. Three machine learning methods (linear, nearest neighbour, and perceptron regressors) have been assessed to decide the best one for this application and machine specifications. Investigations performed revealed good predictive capability of the machine learning model with the multi-layer perceptron regressor. The machine learning-based adjustment of the thermal network exhibited promising results for the potential development of self-tuned thermal models.

Place, publisher, year, edition, pages
2022.
National Category
Electrical Engineering, Electronic Engineering, Information Engineering
Identifiers
URN: urn:nbn:se:kth:diva-321583OAI: oai:DiVA.org:kth-321583DiVA, id: diva2:1711632
Conference
11th International Conference on Power Electronics, Machines and Drives (PEMD 2022)
Note

QC 20221129

Available from: 2022-11-17 Created: 2022-11-17 Last updated: 2022-11-29Bibliographically approved
In thesis
1. Numerical predictions of heat-transfer applied to electrical machines
Open this publication in new window or tab >>Numerical predictions of heat-transfer applied to electrical machines
2022 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

In order to meet the need for increased electrification, and at the same time reduce the total demand for electric energy, behavior change and technological innovation is needed. Over the decades power density of electric motors have increased, leading to increased demands on the cooling system design and performance. The need for reduced energy demand, increased efficiency, and continued increase in performance require continuous development effort regarding cooling systems, understanding of temperature distributions and heat transfer, and thermal simulation tools applicable in the motor manufacturing industry.

    A study on how simulation assumptions affect the resolved temperature field in a traction motor prototype is presented. Here different assumptions regarding loss distributions and air flow distributions are considered. The study illustrates how different simulation assumptions affect the temperature field, and how the results compare to measurements.

    Application of numerical methods for resolving heat transfer, and how the heat transfer is linked to features in the fluid flow, is presented. An air jet impinging on a heated surface is investigated through the application of Large Eddy Simulations (LES) and obtained data processed using the Extended Proper Orthogonal Decomposition (EPOD) method. The study shows the link between structures in the flow and the associated structures in heat transfer. 

    Thermal analysis is an integral part of the motor design and dimensioning process. The method employed in theses studies is often the Lumped Parameter Thermal Network (LPTN). In this work a prototype method for automatic calibration of an LPTN, based on external temperature data, is presented. Application of Computational Fluid Dynamics (CFD) in computing input data needed for LTPNs is presented, where an extension to existing heat transfer correlations related to the end-winding of a form-wound machine is suggested.

    The studies are aiming at enabling advancing the prediction capability of heat transfer and temperature simulation methods applied in analysis of electrical machines.

Abstract [sv]

För att möta behovet av utökad elektrifiering, samtidigt som energibehovet behöver reduceras, krävs både beteendeförändringar och teknologisk innovation. Effekttätheten i roterande maskiner har ökat under decennierna, vilket leder till förhöjda krav på kylningsystemens utformande och prestanda. Kombinationen av behovet av minskning av energibehov, höjd verkningsgrad samt ökade prestandakrav, kräver kontinuerlig utveckling av kylsystem, förståelse av temperaturfördelning och värmeöverföring, samt simuleringsverktyg tillämpbara i motortillverkningsindustrin. 

    En studie av hur antaganden i definieringen av simuleringar påverkar den erhållna temperaturprofilen presenteras i denna avhandling. Studien utfördes med utgångspunkt från en traktionsmotorprototyp. Olika antaganden gällande förlustfördelning och fördelning av kylluftflöde beaktades. Studien illustrerar hur olika antaganden påverkar det simulerade temperaturfältet, och hur resultaten står sig i jämförelse med mätningar. 

    Numeriska metoder för upplösning av värmeöverföring och hur denna är kopplad till strukturer i fluiden presenteras. En luftstråle som infaller på en uppvärmd plan yta studerades med hjälp av Large Eddy Simulation (LES) och erhållen data behandlades med Extended Proper Orthogonal Decompositon (EPOD). Studien visar på kopplingen mellan strukturer i flödet och strukturer i värmeöverföringen. 

    Termisk analys är en viktig del av design- och dimensioneringsprocessen för en motor. En vanligt förekommande metod för detta är termiska nätverk. I denna avhandling presenteras en prototyp till en method för automatisk kalibrering av termiska nätverk, där kalibrering sker mot en extern källa till temperaturdata. Tillämpning av Computational Fluid Dynamics (CFD) för att beräkna in-data till termiska nätverk presenteras också, där en utökning av befintliga korrelationer för värmeöverföring vid lindningsutsticken hos formlindade maskiner föreslås.

    Studierna ämnar till att möjliggöra förbättra predikteringsförmågan hos värmeöverförings- och temperatursimuleringsmetoder tillämpade vid analys av elektriska maskiner.

Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2022
Series
TRITA-SCI-FOU ; 2022:60
Keywords
Electric machines, energy efficiency, heat transfer, thermal management, high fidelity simulation, computational fluid dynamics (CFD), proper orthogonal decomposition (POD), lumped parameter thermal network (LPTN), Elektriska maskiner, energieffektivitet, verkningsgrad, värmeöverföring, värmehantering, simuleringar med hög upplösning, strömningsmekaniska beräkningar, proper orthogonal decomposition (POD), termiska nätverk
National Category
Other Electrical Engineering, Electronic Engineering, Information Engineering Energy Engineering
Research subject
Engineering Mechanics
Identifiers
urn:nbn:se:kth:diva-321585 (URN)978-91-8040-427-3 (ISBN)
Public defence
2022-12-15, https://kth-se.zoom.us/webinar/register/WN_05A23IvPROCKA2beck3VUA, H1, Teknikringen 33, Stockholm, 13:00 (English)
Opponent
Supervisors
Funder
Swedish Foundation for Strategic Research
Note

QC 221118

Available from: 2022-11-18 Created: 2022-11-18 Last updated: 2022-12-08Bibliographically approved

Open Access in DiVA

No full text in DiVA

Authority records

Rönnberg, Kristian

Search in DiVA

By author/editor
Rönnberg, Kristian
By organisation
Chemical Engineering
Electrical Engineering, Electronic Engineering, Information Engineering

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 136 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf