A Data Science Approach for effective Management of Process Safety Incidents

Spread the love

As mentioned in the CCPS Process Safety Guide, an essential element of any improvement program is measure of existing and future performance. The proper analysis of such measures like Leading and Lagging metrices is critical for successful process safety management. Generally, a safety pyramid consists of mix of three types of metrices like Lagging metrices ā€“ which are retrospective set of measures, Leading metrices ā€“ which are forward looking metrices and finally Near-miss which are less severe incidents, which however are very good indicators for future likelihood. Indeed, the availability of data science analysis tools along with enterprise systems is very critical for mitigating catastrophic chemical accidents.

accident action danger emergency
Photo by Pixabay on Pexels.com

The availability of different exploratory methods for assessment of process safety metrices using data science tools like Python will help organizations in effective management of Process Safety Incidents. These analyses are used to carry out the exploratory data analysis of process safety incidents, analysing the lagging indicator relation with Leading indicators and development of safety pyramid structures, Incident Icebergs and Swiss Cheese Models. The Incident Iceberg depicts Fatalities, LWD, Medical Treatments, Minor First aids, Near Misses and Equipment conditions, at Risk behaviours and system weaknesses. Finally modelling barriers like Swiss Cheese Models (Behaviours controls of people) / weaknesses (activities).

Below are the different analytics which can be performed besides modelling complex swiss cheese models. The different data segments which can be used include

  1. Classification of incidents as process safety based on complex criterion like employee, incident location and further analysis using complex regular expressions
  2. Reporting threshold analysis like whether an incident has incurred more than 25K dollars cost due to fires, or explosions, community evacuation reports, involvement of chemicals either flammable or non-flammable to designate as threshold releases or acute releases etc.
  3. Detailed analysis of location where incident occurred and their reporting information like 1-hour release information of chemical exceeding threshold limits or not.
  4. A detailed analysis of incident reporting criteria like process safety damages etc, location where incident occurred, whether planned or unplanned event occurred, an employee Lost time fatality or hospitalization etc.
  5. A successful Process Safety analysis also requires to be analysed for the severity levels and attribution of severity rate calculation for each attributes, like whether they belong to Tier ā€“ 1 or Tier 2 like LOPC incidents
  6. Finally the rate adjusted metrices like Process Safety Total Incident Rate (PSTIR) and Process Safety Incident Severity Rate (PSISR)

 

Besides direct analysis of data points, a detailed analysis of Leading and Lagging Indicator analysis can also be performed. The careful analysis of leading and lagging can help companies identify vulnerabilities and likelihood of incidents. The interrelationship between leading / lagging will help to show the early warning about the integrity of critical controls, likelihood of major accidents, ensuring that companies are properly managing the systems and monitoring them as per requirement and can help in bench marking with other regions and even with peers. To give an example the capture of information around say compliance to monitoring schedule, defects detected with actual occurrence of incidents will give a very good insight. Such details studies can be easily and effortlessly carried out using data science.

Some Enterprise systems already have an option of measuring Process Safety Incidents and capture the information pertaining to process safety information. These systems along with other modules like Management of Change, Audit Management can help in developing plans for overall safety systems for successful process safety management. In reality process safety processes and information availability spans across multiple systems and a early identification of systems consolidation like creation of Data Lakes, which consolidate such information in a single place on a periodic basis will definitely help companies which embark on a long term journey of process safety management. A preliminary careful analysis of process safety indicators for each process area which needed to be captured should not be missed though.

Thanks

Jayakumar

 

 

 

Leave a Reply