Research And Implementation Of Distributed Cloud Platform For Intelligent Design Of Optical Devices

Intelligent design of optical devices is one of the key research areas in the field of nanophotonics.The ever-increasing performance requirements of optical devices demand higher standards of design software.In this thesis,we analyze the fundamental theory of intelligent design of optical devices,and combine emerging technologies such as cloud computing,distributed clusters,and parallel computing to build an intelligent design cloud platform for optical devices.We achieve the entire process of intelligent design of optical devices,from user submission to distributed cluster optimization and simulation and returning results.Then we explore methods to improve the performance of the cloud platform.Our work provides significant reference for the development of intelligent designing software for optical devices.The main achievements of our work are summarized as follows:（1）We have designed and implemented a distributed intelligent design cloud platform for optical devices based on task graphs.In this thesis,we adopted the concept of a task graph to describe the execution order and dependency relationship of various computational tasks in an intelligent design task for optical devices.By analyzing the software and hardware requirements of the cloud platform based on the demands of intelligent design of optical devices and drawing inspiration from mainstream optical device design software,we designed the overall architecture of the platform based on task graphs.From the perspective of business functions,the platform was divided into three modules:user interface,backend control,and simulation-optimization.The user interface module realized functions such as user login verification,optical device modeling and parameters setting,and real-time display of design task progress.The backend control module implements functions such as request processing,task control,database storage and query.The simulation-optimization module implements functions such as iterative optimization,electromagnetic field simulation and distributed deployment.Finally,we validated the functionality of the platform through intelligent design of optical devices,and demonstrated that sufficient cluster resources can improve the efficiency of multi-task execution through a multi-task parallel scenario.（2）Starting from the perspective of improving performance,this thesis proposed improvement methods for the cloud platform.First,the device structure optimization algorithm is split and scheduled based on the task graph,refining the granularity of device design task scheduling.For a complete device design task,it is equivalent to adding "secondary scheduling" and "tertiary scheduling".Through experiments,the improved method had been verified to achieve optimized resource utilization,improved computational efficiency,and better support for load balancing in multi-task operations.Second,the electromagnetic field simulation of multi-input and multi-output devices is parallelized.When solving the electromagnetic field distribution of multiple optical signals,this thesis assigns simulation tasks to multiple computing nodes in the cluster,realizing parallel electromagnetic field simulation of various optical signals,and fully utilizing the advantages of the cluster to reduce the time.The improved method was verified via using Hub₃×33 Hub₄×4 and Hub₆×6 devices,and was shown to have significant advantages in reducing time compared to traditional single-process serial simulation methods.This thesis provides a new approach to solve the problem of large-scale electromagnetic field simulation of optical devices with multiple inputs and outputs,which is computationally intensive and time-consuming.