AI-Driven Cross-Domain Multi-Task Learning for Enhanced Reliability-Based Design with Multi-Fidelity and Partially Observed Data

EasyChair Preprint 14532

Abstract

The integration of Artificial Intelligence (AI) into reliability-based design processes has opened new avenues for optimizing complex engineering systems. This research explores the application of AI-driven cross-domain multi-task learning to enhance reliability-based design by leveraging multi-fidelity and partially observed data. The study addresses the inherent challenges in reliability analysis, including the scarcity of high-fidelity data and the need for efficient computational models that can generalize across different domains. By employing a multi-task learning framework, the research aims to transfer knowledge across related tasks, improving model accuracy and robustness in reliability assessments. The use of multi-fidelity data allows the integration of both high- and low-fidelity models, enabling more efficient resource allocation while maintaining high reliability standards. The methodology also incorporates techniques for handling partially observed data, ensuring that the AI models can make reliable predictions even when complete datasets are unavailable. The proposed approach is validated through case studies in various engineering domains, demonstrating its effectiveness in enhancing the reliability and efficiency of design processes.

Keyphrases: AI-driven, Computational efficiency., Reliability-based design, cross-domain, engineering systems, model accuracy, multi-fidelity data, multi-task learning, partially observed data

@booklet{EasyChair:14532,
  author    = {Kayode Sheriffdeen},
  title     = {AI-Driven Cross-Domain Multi-Task Learning for Enhanced Reliability-Based Design with Multi-Fidelity and Partially Observed Data},
  howpublished = {EasyChair Preprint 14532},
  year      = {EasyChair, 2024}}