Solución de predicción de temperaturas usando datos de un simulador térmico
Cargando...
Fecha
2023-11
Autores
Flores-Jiménez, Yared I.
Título de la revista
ISSN de la revista
Título del volumen
Editor
ITESO
Resumen
Descripción
The industry of integrated circuits is experiencing a moment of fierce change. As is, the methods used in all stages implied in its design process. The present work presents a method to predict temperatures for System on Chip (SoC) chiplet part with quite simple power map and a single thermal interface material using Machine Learning (ML) and its offspring Deep Learning (DL). The SoC part is represented as a response surface of a 2D model geometry surface used for a set of experiments to determine the relevant factors for the temperature prediction. In addition to the experiment design, a deployment strategy to implement a continuous integration and deployment process to be used for the target organization is also proposed. The idea is to achieve the principle of productive ML that states that models should be constantly learning by automating new data ingestion into the training process to enhance model performance in each of the cycle updates. The project proposes a method to strengthen the established thermal processes of the target organization by using ML tools and provide an alternative to speed up thermal model analysis using new available techniques derived from ML and Deep Learning.
Palabras clave
Temperature, Simulation, Simulador, Data Analysis, System on Chip, DoE
Citación
Flores-Jiménez, Y. I. (2023). Solución de predicción de temperaturas usando datos de un simulador térmico. Trabajo de obtención de grado, Maestría en Ciencia de Datos. Tlaquepaque, Jalisco: ITESO.