AI-Powered Safety: Schneider Electric’s Groundbreaking Patent

Schneider Electric announced a pioneering patent acquisition for using AI to optimize the functional safety of industrial processes. The patent involves the use of AI-based reinforcement learning to assess possible process hazards and generate safety/protective mechanisms. The patent introduces a systematic way of conducting hazard analysis. It has the potential to enhance HAZOP (Hazard and Operability) studies by leveraging real-time data analysis and continual learning to adapt and generate multiple possible risks and solutions in simulations.

Using AI to assess and mitigate risk to improve process safety could boost the productivity of industrial facilities and improve worker well-being. Image used courtesy of Pixabay

Building and Retaining Process Safety Knowledge

AI can help build safety hazard knowledge and safety frameworks in industries lacking in this and also retain knowledge lost through industry professionals taking retirement. Concerning the former, AI systems can compile incident reports for use in analyses, predict potential hazards, and generate safety procedures with adaptability, speed, and 24/7 monitoring, previously impossible with oversight solely from human process safety experts.

AI-based hazard analysis systems can also augment process safety expert guidance, enhancing decision-making and helping deliver robust functional safety to industrial facilities and their employees. AI-based systems could also help retain vital process safety data, building a repository from which to learn and adapt new hazard perception and safety models for ever-evolving industrial workspaces. Using AI can eradicate the biases otherwise introduced by safety/hazard assessment teams based on past experiences.

Some might argue that jumping on the AI bandwagon is a sure way to diminish human workers and their capabilities. Others might also say that AI-based inspection, data collection, and analysis concerning worker operations are intrusive. On the other hand, logging and assessing worker incidents could be used as a teaching/learning tool for workers to be aware of their surroundings and maintain good practice and overall well-being.

Schneider’s Pioneering Patent

SE’s patent is intended to improve safety by deploying two AI agents in a replicated industrial environment. The first agent detects possible failures by altering conditions—such as pressure or chemical changes—to cause risks and expose system vulnerabilities.

The second agent minimizes these failures by evaluating preventive measures like alerts and shutdowns. Through repeated simulations, both agents fine-tune their risk mitigation tactics, assuring an information-driven strategy for identifying and mitigating industrial risks before they become actual threats.

As part of the safety instrumented system (SIS), a process safety expert tags process hazards and allocates them to specific protection layers within a facility to program a programmable logic controller (PLC). The safety team writes up a safety requirements specification (SRS), and the safety PLC’s application program is transcribed to comply with the SRS.

The system can design and test a safety PLC program application to make sure that it fulfills safety standards and avoids risks in industrial processes.

Both a Strength and Weakness

In a report from global technology intelligence firm ABI Research, the OT cybersecurity market is projected to rise to $21.6 billion by 2028 from $12.75 billion in 2023 at an approximate compound annual growth rate of 9.2%.

The increasingly connected nature of information and operational technology (IT/OT) in industrial facilities means enhanced communication, improved efficiencies, and data transfer, but also increased vulnerability to cyber threats.

Could AI-programmed PLCs be hijacked by hackers with ill intentions, cause irreparable damage to industrial assets, and endanger people's lives? Image used courtesy of Pixabay

While AI-based hazard analysis and mitigation can enhance safety in industrial facilities, there might also be the potential to critically undermine safety when considering the threat that hackers might pose.

Attackers could hack into safety PLCs and manipulate danger simulations, disable preventive measures, or modify safety criteria, resulting in undetected and dangerous vulnerabilities. This could lead to accidents, equipment breakdowns, or even substantial industrial calamities.

Another significant issue is the hostile exploitation of AI agents. If attackers tamper with the learning process (such as by providing false data or modifying reward functions), the system's capacity to identify and avert risks may be compromised. Without sufficient cybersecurity safeguards, AI-powered safety analysis could become a gateway for cyber-physical attacks, jeopardizing both business processes and human life.

A Hybrid Model Approach, Not Solely AI

SE’s approach to the safe and conscientious implementation of AI in process hazard assessment is to use it as an additional control point rather than a replacement for human supervision and management. The result of this approach is a hybrid model combining human process safety knowledge and AI insights to help businesses onboard AI-based process safety procedures with greater confidence. Using AI in this way can help businesses meet safety requirements, reduce operational costs, and enhance risk prevention for their employees.

Featured image used courtesy of Adobe Stock