Utilization of Artificial Intelligence in Electronics Design
Luukko, Maija (2024)
2024
Sähkötekniikan DI-ohjelma - Master's Programme in Electrical Engineering
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2024-07-12
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202407057522
https://urn.fi/URN:NBN:fi:tuni-202407057522
Tiivistelmä
Artificial intelligence (AI) covers a broad and rapidly advancing field of technology where machines are aimed to perform tasks that would normally require human intelligence. This thesis explores the integration of AI in the design of electronics, focusing on printed circuit boards and integrated circuits. AI technologies are used to automate complex design processes, enhance accuracy, and reduce overall design time which can have large improvements to traditional electronic design methodologies that highly rely on human reliance, simulation, and testing. Implementing AI in electronics design offers numerous benefits, but also comes with new challenges. This thesis aims to explore how can AI be used to help and automate the design of electronics that consists of printed circuit boards (PCB) and integrated circuits (IC), but also what are the new challenges and difficulties that arise from the integration of these AI tools. These questions are answered through both a literature review and practical research, where available AI tools are overviewed and two of these tools are evaluated practically.
The literature review shows promising results for the usage of AI tools in the different phases of electronic design flow. AI technologies aim to overcome the challenges of traditional tools, especially by reducing human dependence by automating processes and increasing both speed and accuracy. AI has successfully automated processes such as PCB routing prediction, translating netlists into physical layouts, and detecting hotspots in IC layouts. With different AI methodologies such as machine learning and deep learning the design processes have successfully sped up and their accuracy and efficiency are enhanced. Also, replacing simulation-based methods with different artificial neural networks improved both the speed and accuracy of methods, but also decreased the computational costs.
However, AI tools rely heavily on extensive data for training and need for computational resources. These results in new challenges which can lead to difficulties with the integration of these AI-based tools with existing software. Transfer learning, optimized neural architectures, and model tuning are possible solutions for both computational power and the need for a large amount of training data.
Multiple companies have started utilizing AI in their electronic design products. However, a lack of clarity about specific AI methodologies remains a challenge. The practical evaluation of the two software leveraging AI shows an offering of potential benefits in circuit design and component selection however the effectiveness depends on user input and careful verification.
The literature review shows promising results for the usage of AI tools in the different phases of electronic design flow. AI technologies aim to overcome the challenges of traditional tools, especially by reducing human dependence by automating processes and increasing both speed and accuracy. AI has successfully automated processes such as PCB routing prediction, translating netlists into physical layouts, and detecting hotspots in IC layouts. With different AI methodologies such as machine learning and deep learning the design processes have successfully sped up and their accuracy and efficiency are enhanced. Also, replacing simulation-based methods with different artificial neural networks improved both the speed and accuracy of methods, but also decreased the computational costs.
However, AI tools rely heavily on extensive data for training and need for computational resources. These results in new challenges which can lead to difficulties with the integration of these AI-based tools with existing software. Transfer learning, optimized neural architectures, and model tuning are possible solutions for both computational power and the need for a large amount of training data.
Multiple companies have started utilizing AI in their electronic design products. However, a lack of clarity about specific AI methodologies remains a challenge. The practical evaluation of the two software leveraging AI shows an offering of potential benefits in circuit design and component selection however the effectiveness depends on user input and careful verification.