Aligning Predictive Maintenance with Safety Instrumented Systems
Key Highlights
- Safety instrumented systems should never depend on predictive maintenance data for critical decisions, with human oversight required for all safety-affecting actions.
- Companies must audit systems against IEC 61508/61511 standards while ensuring predictive maintenance recommendations inform but never override safety protocols.
- Segregated OT networks, hardened gateways, encrypted industrial protocols and fail-safe fallbacks prevent predictive maintenance connectivity requirements from degrading safety functions.
In the industrial automation sector, a variety of software technologies exist to address nearly every aspect of operations. Sometimes, however, those systems can come into conflict. Consider the case of predictive maintenance systems and safety instrumented systems.
Predictive maintenance (PM) systems monitor machinery and other equipment in real-time to identify potential failure points before they become real failures. Typical PM systems employ diverse sensors to track critical parameters such as machinery wear and tear, temperature changes, vibration and shock, pressure and fluid levels. This enables technicians to perform proactive maintenance to minimize and, in some cases, eliminate downtimes and maintain efficiency.
Other features of PM systems include integration via Industrial Internet of Things (IIoT), which enables remote monitoring and control. With this capability, PM systems can incorporate cloud-based artificial intelligence (AI) to collect, analyze and process massive volumes of data in real-time to help manufacturers boost productivity and extend asset lifespans.
Safety instrumented systems (SIS) actively protect personnel from injury, as well as equipment from damage. Traditionally, a SIS consists of hardware and software controls capable of shutting down a factory operation or operations when the SIS detects a hazardous condition. Depending on their SIL level, some SIS types include emergency shutdown systems, ventilating systems that divert hazardous gases to prevent explosions and toxic environments, and interlock systems that prevent processes from resuming operation until maintenance personnel can correct the fault.
Issues with functional overlaps
To ensure compliance with safety standards, such as safety integrity level (SIL) standards SIL2/SIL3, IEC 61508 for functional safety and IEC 61511 for safety instrumented systems, a SIS will continue to operate correctly even though a system fault and/or shutdown occurs.
Whereas a PM system detects potential faults that can be resolved proactively, a SIS detects actual fault occurrences and executes the proper actions to prevent personnel injuries and damage to physical assets.
Although PM systems provide significant benefits, they may conflict with SIS protocols when it comes to fail-safe operation. That’s why a SIS must operate independently of all other control systems including PM systems to prevent any compromises to its safety functions.
When it comes to having a PM and SIS operating in the same area, there are several critical issues manufacturers need to consider, such as how to safely integrate a PM system alongside an existing SIS, maintain compliance with safety standards, choose the proper platforms and protocols, and ensure reliable data integrity and connectivity.
To get insights into the coordination of these systems, Automation World connected with Hugo Dozois-Caouette (HDC), chief technology officer and co-founder of MaintainX, a supplier of AI-powered maintenance and asset management software.
AW: How can manufacturers and OEMs integrate PM systems with their existing safety systems without compromising deterministic response times required for reliable fail-safe operations?
HDC: Preventative maintenance systems should always be both complementary and additive to safety instrumented systems and not replace them. SIS controls should stay local to the machine with fixed controls and rules as to when they should trigger. Rather than the maintenance team replacing any critical gates on machines, PM systems should empower the team by helping them get ahead of breakages and other issues.
AW: What safeguards should PM analytics platforms integrate to ensure maintenance optimization recommendations never override or interfere with critical safety functions?
HDC: Safety functions should always come first and be able to override what the PM system is doing. The SIS must work completely independent of the PM system. And rather than merely apply automated system recommendations blindly, it's also important to keep a human in the loop as another layer of safety. The usual audit trails for recommendations and signoffs are highly critical, especially if they can affect safety.
AW: How do PM technology vendors validate or ensure predictive algorithms will not create any false positives that may lead to unnecessary safety system bypasses or maintenance interventions during critical operations?
HDC: Vendors should validate their models on historical data and run them in both passive and shadow modes in the plant during active operations. They should then generate surface confidence scores and require multi-sensor corroboration. And, again, they must keep a human in the loop to monitor operations and any action that can affect safety behavior.
AW: How can PM suppliers maintain compliance with safety standards like IEC 61508 for functional safety and IEC 61511 for safety instrumented systems while implementing predictive maintenance capabilities?
HDC: Systems that are bound by both standards should be audited and controlled to ensure their compliance. They should also implement connections to PM systems in ways that are non-blocking. In other words, the SIS should not depend on the PM system for information.
Having clear boundaries between systems is critical. PM systems should reliably and consistently inform the maintenance team as to what events may be coming and that information can then be surfaced within the SIS. However, in the end, the SIS is the sole source of truth and needs to be integrated so as to not depend on any external information that may not possess all the necessary safety certifications.
AW: What protocols are viable for managing situations where PM data suggests equipment optimization that could conflict with safety system operating parameters?
HDC: While the PM system should aim for ingesting safety requirements and integrate those in its recommendations, the PM system cannot ensure it has all the required safety information to guarantee its recommendations. That’s why SIS requirements must always take precedence over the PM system.
AW: How can companies ensure data connectivity and communication systems for predictive maintenance do not introduce new failure modes and/or cybersecurity vulnerabilities that could compromise safety system reliability?
HDC: Companies should always design their systems with a defense-in-depth approach. They should segregate operational technology (OT) networks, use hardened gateways, and encrypt and authenticate industrial protocols using technologies like mutual transport layer security (TLS) for MQTT and OPC-UA communications. Additionally, enterprises must enforce least-privilege and short-lived credentials, as well as ensure redundant, fail-safe fallbacks so any connectivity loss cannot degrade SIS functions.
As an example, MaintainX’s Asset Connect approach explicitly uses MQTT and Sparkplug for PLC ingestion while recommending hardened gateway patterns for industrial environments. The Asset Connect platform lets OEMs proactively monitor their original equipment installed across their customer’s facilities wherever they are located.
More predictive maintenance and safety insights from Automation World:
About the Author
Mat Dirjish, contributing writer
Mat Dirjish has years of tech reporting experience at B2B technology publications such as Electronic Products Magazine, EE Product News, Electronic Design, Sensors Online & Sensors Expo, and Sensors Daily. He's been a regular contributor to Automation World since 2023.
