Advanced AI-Based Real-Time Industrial Safety Sentinel for Smart Hazard Detection and Workplace Safety
Snehaprabha Jadhav1, Yogini C. Kulkarni2, Pramod Jadhav3, Vinod Patil4, Amol Kadam5
1Snehaprabha Jadhav, Assistant Professor, Department of Information Technology, Bharati Vidyapeeth (Deemed to be University) College of Engineering, Pune (Maharashtra), India.
2Prof. (Dr.) Yogini Kulkarni, Assistant Professor, Department of Information Technology, Bharati Vidyapeeth (Deemed to be University) College of Engineering, Pune (Maharashtra), India.
3Dr. Pramod A. Jadhav, Associate Professor, Department of CSBS, Bharati Vidyapeeth (Deemed to be University) College of Engineering, Pune (Maharashtra), India.
4Dr. Vinod H. Patil, Department of E&TC Engineering, Bharati Vidyapeeth (Deemed to be University) College of Engineering, Pune (Maharashtra), India.
5Dr. Amol K. Kadam, Associate Professor, Department of CSBS, Bharati Vidyapeeth (Deemed to be University) College of Engineering, Pune (Maharashtra), India.
Manuscript received on 01 April 2026 | First Revised Manuscript received on 20 April 2026 | Second Revised Manuscript received on 03 May 2026 | Manuscript Accepted on 15 May 2026 | Manuscript published on 30 May 2026 | PP: 18-25 | Volume-15 Issue-1, May 2026 | Retrieval Number: 100.1/ijrte.B836615020726 | DOI: 10.35940/ijrte.B8366.15010526
Open Access | Editorial and Publishing Policies | Cite | Zenodo | OJS | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open-access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Industrial environments — including factories, construction sites, warehouses, and chemical plants — continue to experience hazardous incidents due to PPE noncompliance, unauthorised zone entry, and unsafe proximity to workers. Simultaneously, web, IoT, and edge applications deployed in these environments remain vulnerable to well-documented cyber threats, including SQL Injection, XSS, and broken access control. This paper presents the Intelligent Integrated Cyber-Physical Safety and Security Framework (IICPSSF) [11], a novel hybrid edge-cloud AI system that uniquely and simultaneously enforces (i) real-time vision-based physical industrial safety monitoring, and (ii) adaptive cybersecurity design pattern enforcement — governed by a shared Unified LLM Assisted Natural Language Rule Engine (ULNLRE). The edge layer employs YOLO-E, an open-vocabulary object detection model, for promptable real-time perception at approximately 60 FPS. At the same time, a deterministic symbolic rule engine enforces auditable safety and security policies. A Pattern Knowledge Base and context aware adaptive selection engine handle cybersecurity pattern recommendations for web, IoT, and edge applications. The system achieves 92.5% precision, 90.7% F1-score, and 99.5% specificity on physical safety violation detection across four violation types, and demonstrates effective coverage of six OWASP Top 10 vulnerability classes.
Keywords: Cyber-Physical Security, Industrial Safety Monitoring, PPE Detection, YOLO-E, Edge-Cloud Computing, Rule-Based Reasoning, Explainable AI, LLM Policy Translation, OWASP, Secure Design Patterns.
Scope of the Article: Computer Science and Engineering