Advancing Process Hazard Analysis: AI Applications in Design and Revalidation HAZOP
Where to start AI journey
The HAZOP study will be performed during the Front-End Engineering Design (FEED) stage and the detailed Engineering Stage. Also, as per OSHA and other regulatory bodies, the HAZOP study must be revalidated every five years.
During the design stages, Hazard and Operability (HAZOP) studies are conducted with the participation of licensors, contractors, vendors, project management consultants (PMC), and company employees. This ensures that the design is thoroughly reviewed and the available data is validated. Companies can achieve quality HAZOP studies if they are ready to spend money and schedule appropriately.
However, During Revalidation, Operations, Maintenance, Engineering services and Technical services teams are required to evaluate the current operation and track changes in the plant.
Sparing key personnel for Revalidation HAZOP while the plant is running creates a challenging situation for the management. Hence, we urge to
“use the help of AI for PHA revalidation to overcome the unavailability of key personnel.”
The authors conducted a thorough analysis by considering parameters like the urgency of the requirement to automate, Availability of AI tools and quality data, training of the AI module and validation of AI results, with the objective of deciding the starting point for using AI in Process safety studies.
In this analysis, digitalized plant data is assumed to be available. For the selected aforementioned parameters, quantitative evaluation is performed (refer to Table-2) for using AI in Design HAZOP and Revalidation HAZOP.
Table 2 Evaluation for AI implementation
*% is the authors view based on experience
From the above analysis, it is recommended that Revalidation HAZOP be the best starting point for using AI in process safety studies.
‘Recommend to start the AI journey in PSM studies from revalidation HAZOP’
2. Plan of Action for AI in revalidation HAZOP:
In addition to the quality-labelled data and trained personnel, the following activities shall be performed to achieve better benefits.
a) Integration of existing Plant data
b) Select quality HAZOP report to train the AI modules
c) Start with Rule-based systems
d) Move to the advanced AI Algorithms
a) Integration of existing Plant data:
Many companies have started integrating and accumulating data in one place for various benefits. AI implementation may demand a few more integrations, like MOC Records and Inspection Reports, to be reflected on the Intelligent P&ID. More Quality and updated information yields accurate results.
Inputs (from the digitalisation) have been significant in giving inputs to the HAZOP worksheet. For example,
1. Changes in the P&ID and findings in Inspection & Corrosion Reports will help us identify the new causes and consequences.
2. DCS trends, alarm history will help us revalidate the safeguards.
Based on these changes in the Causes, Consequences and Safeguards, the Risk ranking will be modified as depicted in Figure 7.
