Overall strategy for the implementation of AI in PSM Studies
The overall strategy for extending AI implementation to all the PSM Studies will be highlighted in this section. Key steps involved in overall strategy are summarized below:
a) Digitalization
Digitalization: Data is crucial for AI. Every industry or processing plant requires its own historical data to develop and train AI when needed. Even if there are no immediate plans to implement AI, it is recommended to begin digitization processes such as Smart P&ID, centralizing DCS data, maintaining records, and incident investigation. Process Digital Twin can also be a valuable tool. By preparing in advance, you can quickly implement AI when the time is right.
b) Transitioning from rule-based to machine learning and deep learning:
1. The first rule of AI implementation in the industry is to start with Rule-based implementation. For example, we can use AI to revalidate the safeguards and consequences based on certain rules, as explained in Section 5.
2. When the rules become complex, upgrade to a simple model. Using simple Machine Learning Algorithms, we trigger AI to think of certain operation concerns like Off spec, identifying further causes etc.,
3. When your model reaches its limit, upgrade to a more complex one. To perform further in-depth analysis, we can employ sophisticated algorithms like Random Forest and Deep Learning.
c) Utilizing Explainable AI
One of the most significant issues with AI today is the lack of trustworthiness. Although we receive results from AI, we often don’t understand how those results are determined. For this reason, a new field of research has emerged called “explainable AI” or XAI. The XAI report aims to explain how AI results are obtained, which factors are considered, and which are not.
XAI will assist in making an informed decision. Additionally, it helps in addressing legal arguments. Hence, it is important to consider a vendor’s XAI capabilities when selecting an AI implementation.
d) Building a skilled internal team
To implement AI on a large scale, it is crucial to have a competent internal team that can handle the technical, operational, and ethical challenges of AI. Building such a team can be a challenging but rewarding task. The process involves several steps, including:
-
Assessing internal capabilities, including data availability, talent pool, infrastructure, and culture.
-
Building skills in the workplace, such as identifying skills gaps, targeting learning to roles and individuals, and more, is critical to success.
-
Consider setting up an AI centre of excellence, which is a dedicated unit that has the goal and vision for implementing AI, identify new case studies, manage external partners, and share best practices.
This approach can help ensure that your AI initiatives are strategic, efficient, and effective. Suggested process for building a skilled team is depicted in Figure-13
e) Fostering collaboration between Human Intelligence and Artificial Intelligence
Human and artificial intelligence should work together to reap the maximum benefits. Table 5 compares the strengths of Human and artificial intelligence.
Table 5 Comparison of Human and Artificial Intelligence
e) Conclusions
The use of AI can greatly enhance process safety studies, provided that proper digitalisation and high-quality data are available. To start this journey, it is recommended to employ AI in a Revalidation HAZOP study, which allows process industries to utilise their resources optimally. Eventually, this approach can be extended to other PHA techniques as well.
Developing a competent internal team and collaborating human intelligence with AI through Explainable AI will maximize the potential for success of implementing AI in process safety.
References
The use of AI can greatly enhance process safety studies, provided that proper digitalisation and high-quality data are available. To start this journey, it is recommended to employ AI in a Revalidation HAZOP study, which allows process industries to utilise their resources optimally. Eventually, this approach can be extended to other PHA techniques as well.
Developing a competent internal team and collaborating human intelligence with AI through Explainable AI will maximize the potential for success of implementing AI in process safety.