AI-Driven Adaptive Training Architecture for Post-Disaster Structural Damage Assessment

Abstract

Post-disaster structural damage assessment demands rapid, consistent, and regulation-aligned decision-making under uncertainty, yet traditional training approaches lack individualized progression tracking and performance-aware feedback. This study proposes an AI-driven adaptive training architecture that integrates modular domain-specific content with a competency-based learner modeling framework and a structured analytics pipeline. Learner interactions - accuracy, response time, error patterns, and repetitions - are transformed into dynamic mastery estimates across predefined competencies. An adaptive recommendation engine uses these states to generate personalized remediation and next-step guidance. The layered architecture ensures scalability, maintainability, and regulation-aware content updates. A formal mathematical representation supports transparent competency tracking and real-time updates. Preliminary pilot deployment demonstrates stable mastery progression, consistent weak-competency identification, and short-term performance improvements following adaptive recommendations, with operational stability under concurrent use. This work bridges post-disaster assessment methodologies with adaptive learning systems, contributing a scalable, performance-driven training infrastructure for safety-critical professional education. Future work includes controlled experimental validation and statistical evaluation of learning gains. © 2026 IEEE.