Ignoring your data is Crowcon’s seventh in the series of Deadly Sins of Gas Detection. A recent news story about an oil worker found collapsed over an open hatch, dead, highlighted this all too graphically. One of the most tragic aspects of this story (tragedy being properly defined as something that could have been prevented) was that data which could have saved him was logged in his personal gas detector.
An opportunity missed…
Three weeks before, while sampling from a similar hatch, this same man had been taken ill and sought medical assistance. His symptoms subsided and he was discharged, undiagnosed. Later, data logs retrieved from his gas detector showed that he had experienced a 5 minute period of low oxygen while he was working at the hatch. Oxygen below 19.5% is considered deficient – he experienced levels as low as 10%. On the day he was found collapsed and dead, his detector showed oxygen levels had fallen to 7%.
It seems reasonable to speculate whether this worker ignored a gas detector alarm – see previous Deadly Sin. Regardless, had the original incident been investigated, and the low oxygen levels observed in the data in time, suitable remedial action could have avoided the oil worker’s death. But the data was not reviewed until too late, so the opportunity to save him was missed.
Spot the Trends
It is becoming common for fixed and portable gas detectors to gather and store data from their immediate environment. Use of this gas detection data to spot trends and improve safety is in its infancy for many. In other industries, such as food and pharmaceutical production, analysis of data gathered from the working environment is common practice. Data analysis, and the information that can be gleaned from it, guides the setting of sensible alert and alarm limits for safety critical factors.
Know your limits
By plotting the data gathered from monitoring, the operational norms for an area or process can be recognised. These norms reflect the routine operational conditions. If the conditions move outside this range, something must have changed. Alert limits can be set around these operational norms. Once established, “outlier” data points, which occur outside these alert limits, trigger an investigation. By investigating at the alert stage, the chance of a situation escalating into a full blown alarm level incident can be avoided.
Identify the patterns
An erratic spread of data points could indicate that some aspect of a process is not under control. By plotting the data against time and sometimes space, patterns can be identified. It may possible to establish whether a particular issue can be associated with:
- An event – such as a shift change or the turning on or off of a piece of the plant machinery
- A location – for instance, by vents or hatches, which can develop a different microclimate that results in outliers
- An individual – due to a particular aspect of their role (sampling at hatches, for example) or their own behaviours and practices.
This list is not exhaustive, but it illustrates how data can be used to identify patterns that could be used to predict incidents and so prevent them before they arise.
Prevention better than cure
The ease of storing and retrieving data logged on modern fixed and portable gas detectors is likely to raise the expectations and demands of health and safety authorities on companies to employ these functions to improve safety. Use of data analysis in gas detection to protect staff can only increase as the predictive and preventative powers are increasingly recognised.