Unlocking Industrial Intelligence: Smart RCFA Using Adaptive Machine Learning and Feature Selection on Complex Multi-Modal Input Multioutput Data

The complexity of modern industrial systems demands intelligent and adaptive approaches for effective Root Cause Failure Analysis (RCFA). This paper introduces a smart RCFA framework designed to process and analyze complex multi-modal input multi-output (MMIMO) data in the form of numeric signals, visual imagery, and descriptive text. By integrating adaptive machine learning with advanced feature selection techniques, the proposed system identifies and prioritizes critical features across heterogeneous data types, enabling accurate classification and interpretation of failure causes. The framework employs multimodal learning strategies to fuse and extract meaningful patterns from structured sensor data, image-based diagnostics (e.g., thermographic or visual inspection), and contextual descriptions such as maintenance logs or operator reports. Experimental validation using realworld industrial datasets demonstrates the model's ability to deliver high diagnostic accuracy and actionable insights, even in environments characterized by data diversity and complexity. This work represents a significant step toward unlocking industrial intelligence through holistic, data-driven failure analysis, supporting predictive maintenance and informed operational decision-making.