MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202621064756 A) filed by Mr. Chander Vijay S Sanbhi on May 22, 2026, for Cross-Domain Reliability Knowledge Graph With Transfer Learning For Multi-Industry Failure Mode Intelligence..
Inventor includes Mr. Chander Vijay S Sanbhi.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: ABSTRACT [505] Modern industrial reliability engineering faces a foundational knowledge fragmentation crisis wherein failure mode intelligence—comprising root cause mechanisms, progression dynamics, environmental sensitivities, and mitigation strategies—remains siloed within individual industry verticals including aerospace, automotive, energy, maritime, semiconductor manufacturing, and medical devices. Consequently, a bearing spallation mechanism learned from offshore wind turbine gearboxes at significant cost cannot inform predictive maintenance models for mining conveyor rollers, even when the underlying physics of failure are nearly identical, resulting in redundant investigative expenditure, persistently high false alarm rates, and preventable catastrophic failures across global industrial infrastructure. [510] Existing reliability knowledge management systems exhibit critical deficiencies in their capacity to represent failure modes as machine interpretable graph structures spanning multiple industries, automatically discover cross domain analogical transfer opportunities where similar physical mechanisms manifest in superficially different components, adapt predictive models from data rich industries (e.g., automotive fleet telemetry) to data poor industries (e.g., nuclear facility auxiliary systems) without compromising safety certification boundaries, and continuously ingest unstructured maintenance narratives, repair logs, and incident reports to expand the knowledge graph without manual ontology engineering. [515] The convergence of heterogeneous knowledge graph architectures, domain adversarial neural networks for invariant feature learning, semantic similarity embedding spaces, and graph neural network based transfer learning presents transformative opportunities for unifying industrial failure mode intelligence. Systems capable of learning failure mechanism embeddings that transcend industry specific component taxonomies can autonomously transfer reliability insights across industries, dramatically reducinwg dthaeta “cloold start” problem for noorv real re assets, and enabling proactive failure prevention before first incident occurrence in a target domain. [520] The present invention describes a comprehensive Cross Domain Reliability Knowledge Graph with Transfer Learning for Multi Industry Failure Mode Intelligence. The system integrates: a heterogeneous reliability knowledge graph spanning 28 industrial taxonomies with 2.3 million annotated failure mode nodes, causal relationship edges, and mitigation strategy edges; a domain adversarial neural network that learns failure mechanism embeddings invariant to industry specific surface features; a semantic transfer learning engine that identifies candidate cross domain analogies using hyperbolic graph embeddings; and a continuous ingestion pipeline that parses unstructured maintenance narratives using fine tuned large language models with failure mode entity recognition. [525] Validation studies conducted across six industry partners—including commercial aviation, onshore wind energy, semiconductor fabrication, and heavy mining equipment—demonstrated that the platform achieved 84.7 percent accuracy in predicting failure modes for previously unseen asset types in the target industry without any target domain training data (zero shot transfer), 63.9 percent reduction in false positive alarm rates when adapting models from automotive to maritime diesel engines, 72.4 percent improvement in root cause identification speed for novel failure incidents, and discovered 147 previously undocumented cross industry failure analogies validated by domain experts, including a pump cavitation signature that successfully predicted turbine inlet valve failure patterns in hydroelectric plants. [530] The research findings confirm that the Cross Domain Reliability Knowledge Graph constitutes a foundational technological advancement for multi industrial reliability intelligence infrastructure, with deployment potential spanning industrial asset operators, reliability consulting firms, original equipment manufacturers, predictive maintenance software vendors, and insurance underwriting organizations requiring unified, transferable, and continuously learning failure mode intelligence across diverse industry verticals.
Disclaimer: Curated by HT Syndication.
