The future of bowties: predictive risk management?

^{Figure 1: Basic bowtie diagram example. Click here to view full image.} 

These bowties (such as Figure 1) are typically at an abstraction level where the barriers do not have one specific counterpart in reality. In other words, a Pressure Relief Valve (PRV) as a barrier on the bowtie does not point specifically to one actual PRV that is physically present in the plant. It is more like a label that accounts for all PRVs in for that particular process (hazard) that have an effect on that scenario line.

After creating the initial bowtie diagram, a common next step is to add meta-data to bowtie elements (see Figure 2). This helps to understand more about the scenarios and barriers and adds depth to the visualization of risk. It also helps to uncover common failure modes and aids the identification of the presence or absence of variation in barrier types. In addition to that, assigning job titles to barriers and supporting activities also helps with enforcing barrier ownership and accountability.

^{Figure 2: Bowtie with meta-data. Click here to view full image.}

Dynamic bowties

Having a bowtie diagram with meta-data does not have to be the end-point. To help making risks understandable, bowties can be used for (operational) barrier management. To go from static bowties into dynamic bowties they have to be fed with real-life information. Two often used sources of information are incident data and inspection/audit data (as shown in Figure 3). By displaying the data from risk-based audits and barrier-focused incident analyses on bowtie barriers, the bowtie comes alive and starts to reflect the actual performance status of the risk control framework. Performance trends can be spotted and focused improvement actions can be undertaken.

^{Figure 3: Part of a bowtie with performance data. Click here to view the image.}

Bowties that are fit for this purpose are often different from the bowties that were used for risk analysis sessions only. The abstraction level comes down a bit, to be able to map incident barriers to bowtie barriers and to ask specific audit questions about them.

Real-time performance

A next step in barrier management is to operationalize bowties even more so that they would give an almost real-time representation of the status of (safety) systems. This would ask for the linkage of various data sources (e.g. maintenance data, permit data, competence data) that are relevant for barrier performance.

To achieve the above, the bowties have to be adapted further, to match the physical reality and be able to map the data. An example would be the maintenance data, where the Computerized Maintenance Management System (CMMS) asset tags have to be represented on a barrier component level.

There are choices to make here, based on the required level of detail the organization needs:

1. A bowtie barrier is actually a collection of real-world components with the same function (Figure 4). There are thresholds or ranges where the barrier state is online, offline or degraded. The high-level barrier state represents the fact that ‘too many barriers/components in this group are offline/degraded’.
2. A bowtie barrier has a one-to-one counterpart in reality. The barrier on the bowtie is Pressure Relief Valve 010.23.4200A and that PRV is physically present in the facility.

When opting for number 2, the number of bowties would need to increase to represent all critical processes. Number 1 gives an indication of which type of barrier is degraded, but you would still need to drill one layer deeper to find the affected barrier and process.

^{Figure 4: Bowtie with linked CMMS data. Click here to view the full image.}

Linking hardware component data from CMMSs is one possible dataset and already fully supported in BowTieXP. Recent developments are steering towards being able to map more data sources and thus visualizing bowties suitable for data-driven day-to-day risk-based decision-making.

^{Figure 5: Online status overview of a barrier system. Click here to view the full image.}

Predictive risk management?

Nobody knows exactly what the future holds, but an intriguing thought is to use bowties and barriers to predict the short- and long-term future of risk levels in an organization, by mapping historical and real-time data. If process sensor input (e.g. for pressure and temperature) is linked to bowtie threats, barrier status data (e.g. maintenance, competence, permits) to barriers, and that data is combined with historical data, would it be possible to accurately predict where potential problems will arise? Will we know exactly where to spend our resources first, to improve the scenarios that need our attention most? Dare we trust in what the predictive bowtie picture is telling us? With the broader developments in the fields of big data and technology in general, it is not hard to imagine how this could complement and evolve barrier management even further.

If you have thoughts on this subject that you would like to share with us, please contact us.

© CGE Risk. 2018 – The copyright of the content of this blog belongs to CGE Risk Management Solutions B.V