Minimizing Steam Eruptions during coke cutting in Delayed Coker Unit to improve personal safety.

About Customer

  • Customer is a 15 million TPY Refinery operator in India. Refinery consisting of 36 plants including Delayed Coker unit.
  • The customer had faced issues of Health Safety and Environmental due to Geysers / Steam eruptions during Coke Cutting step in their Delayed Coker Unit operations.
  • Geysers/eruptions: Under abnormal or anomalous operating conditions during steaming or quenching steps, short-circuiting of channels in coke bed happen leading to hot spots in the coke drum.
  • When the drums are taken out for coke cutting (i.e the process to cut coke using water jets), water comes in contact with isolated pockets of hot coke, resulting in a geyser of steam, hot water, coke particles spread across 2~3 miles of plant radius.
Process Point - Customer

Objective / Challenges

  • Customer technical services team tried to solve the problem using domain expertise and descriptive data analytics.
  • They were not able to correlate factors that induce steam eruptions and unable to visualize key indicators.
  • This is due to the following reasons
    • A lot of explanatory parameters (Over 80+ operating parameters)
    • The volume of data was very high to analyze
    • Building correlations between parameters was not possible as they were highly interdependent.
  • Due to the above challenges, a customer had engaged us to use advanced analytics to solve the problem.

Solution to problem

  • We have used cognitive analytics and data sciences to correlate a large amount of plant data across several operating parameters consisted of 60 measured variables.
  • Multivariate analysis was used to contextualize a large volume of data sets.
  • Predictive Algorithms were developed using various classification, clustering, and regression algorithms to predict eruptions accuracy.
  • Developed a predictive model and converged it with the observed eruptions based on historical data provided by the client.
  • The model has predictive power with an accuracy of 95%.

Benefits realized

  • Cognitive analytics and data sciences helped to correlate a large amount of plant data across several operating parameters.
  • Based on predicted potential steam eruption scenarios:
    • Operations could undertake corrective actions and
    • The  Adoptive Intelligence model is being deployed in the plant facility for real-time alerts to warn ahead of time for any potential safety issues.