All production facilities will experience downtime sometime during production, where the facility will fail to operate within safe limits. The root causes or reasons behind this is either technical, operational, or organizational.
Can downtime be avoided by the control room operator with the use of new digital solutions?
Let us compare the operator in a control room to a formula one race car driver. The overall performance will be dependent on both the technical solutions in the car, how well the support team is organized, and how well the driver operates (drives) the car. Today, we see lots of new safety features in a car assisting the driver. Similarly, digitalization can provide information assisting both the control room operator and the operational support team.
How much trouble does the lack of situational awareness cause?
The Abnormal Situation Management Consortium with members from e.g. BP, Total, ConocoPhillips, ExxonMobil, Shell, Sasol, and Honeywell, have completed some interesting studies showing how industrial plants will lose 3-8 % of production due to unplanned upsets.
Additionally, 50% of unplanned upsets occur due to a lack of situational awareness. 90% of these events are preventable. The sources of these abnormal events are illustrated in figure 1 below.
Past attempts to help the operator
So, what helping tools have we provided the operator with to understand the situations?
A typical control room operator will have around 3-400 process graphics and 30-40 000 configured alarms to assist them with the decision on how to respond to a plant upset.
Typical user control room operator HMI
With more information through IoT (Internet of Things) and digitalization initiatives, we must ensure that the added information and algorithms do not add more complexity to the situation.
Past attempts have improved the physical design of the control room environment to ensure that the working conditions are satisfactory. IFE, Halden has done a lot of work in this area and supplied extensive design tools, such as HVRC, to facilitate the iterative design process according to Human Factor Engineering standards, such as ISO 11064. Managing and optimizing the alarm systems is also a valuable contribution to ensure that control room operators are not overloaded.
Is digitalization really helpful?
We currently have a new set of tools within digitalization, covering IoT, where we can add plenty of additional sensors and get more information from each existing sensor through communication protocols. Data analytics or “Big Data” can be used for creating new advanced analytic algorithms, which provides better insight. The concept of machine learning from these huge amounts of data should enable us to better understand the situation and to improve situational awareness. Some people even suggest replacing the control room operator with Artificial Intelligence (AI). However, they might not have a complete understanding of the challenges within AI. The HSE implications need to be considered more carefully and control room operators will still be needed in the future. Hopefully, we can develop new tools that will help the operator in becoming more efficient and improve the decisions made by giving them a better understanding of the information they receive, and not applying more information points.
The limitations of using Artificial Intelligence
There are a few limitations we need to understand when applying data science and machine learning.
Presenting the results of the data analytics and learning as a black box calculation will not build trust in the results.
We are still not close to replacing the human in the decision loop for complex assets, the results should be presented as decision support.
You will experience that the system suggests that something is faulty when it is not (known as a False positive) and that the system does not recognize anything when there is a fault (known as a False negative).
Engineering knowledge and feedback is needed. The people that know the facility best will be needed to assess how trustworthy the results are, and to provide a closed feedback loop to remove faulty suggestions.
The system will not be good at detecting first-of-a-kind incidents.
What to consider before applying new technologies
Several considerations need to be covered when introducing more digitalization in the control room or the process itself:
Information overload, reduce data points, and provide more information
Involve the user in the development. People do not believe in black box results.
Show the operator how the results have been obtained.
Do not trust the new technology blindly – machine learning will also learn how to make mistakes!
Use both explicit and tacit knowledge to verify results.
We want to support the control room operator through the cognitive decision process (Yingxu Wang, 2007). This should involve two phases: analysis and action. The analysis is involved to notice deviations, understand the root cause, and predict further consequences. The action involves action planning based on available means and given objectives. Recommended reading for a better understanding of this topic is “Thinking Fast and Slow” by Daniel Khaneman (Farrar, Straus and Giroux, 2011).
Strong opinions with little to no facts
Humans often avoid the cognitive decision process, because it is unnatural for us. We are quick to make decisions based on little experience or facts. This thinking method is useful when driving a car, but can be dangerous in more complex, high-consequence situations, like operating offshore installations. The way you use your brain while working on a difficult mathematical assignment - that’s what we're looking for. Providing control room operators with a better understanding of the information given when a difficult and stressful incident occurs is crucial to ensure safety and reduce downtime. Maybe the answer to the difficult mathematical problem should be - why not give the operators a calculator?