Advanced Analytics for Process Safety

A look at how advanced analytics facilitates and improves process safety efforts within the chemical process industries (CPI)

Simply by the nature of some of the chemicals used and the associated complex processes involved, chemical plants can be hazardous. With the potential exposure to toxic substances, fires and explosions, incidents can range from minor to extremely serious. They can also result in significant plant damage, injuries, and sometimes even loss of life. Furthermore, major chemical manufacturing incidents can also have a catastrophic impact on the surrounding community and environment. Some examples of serious chemical accidents in recent years include the following:

• The Williams Olefins Chemical Plant in Geismar, Louisiana in 2013 [1]— A reboiler rupture resulted in a vapor cloud ignition, due to an unknown source of fire or heat that resulted in two fatalities
• Chevron Oil Refinery in Richmond, California in 2013 [2] — Personnel accidentally triggered a spark as they worked to remove insulation from an extremely old pipe that was already leaking crude oil, which resulted in a catastrophic explosion
• Danlin Chemical Plant in Thomas, Oklahoma in 2013 [3] — A fire from an unknown source engulfed the plant, which caused explosions of pressurized containers
• Jiangsu Tianjiayi Chemical Plant2 in Yancheng Jiangsu Province, China in 2019 [4] – A fire from an unknown source spread to pesticides storage containers causing an immense explosion
• Philadelphia Energy Solutions Complex in Philadelphia, Pennsylvania in 2019 [5] — The failure of a corroded pipe resulted in a fire that triggered three massive explosions and the release of 5,239 lb of toxic hydrofluoric acid

Although there can be numerous potential causes for chemical industrial incidents, one cause is equipment failure. The industry continually faces challenges to ensure proper plant process safety and reduce the risk of incidents. Additional challenges facing the industry include improving plant efficiency, increasing product quality and production yield, and reducing energy consumption. The chemical process industries (CPI) can achieve all of these, along with improved process safety efforts, through the adoption and use of advanced self-service analytics.

A new era of safe processing

CPI plants today capture an immense amount of process data, all of which can — and should — be leveraged to better understand chemical processes. Traditionally, when incidents occur, global process experts and data scientists are often brought in. They will then “deep dive” into historical time-series data and construct mathematical models to help local production teams better understand what happened and determine how they can prevent future incidents. This approach can be inefficient, however, as there are rarely enough global experts and data scientists for the amount of work needed. It is also costly and time-consuming.

A more viable and efficient approach would be to equip process experts with the tools to perform the analytics themselves. This is where advanced self-service analytics comes in play. This tool brings significant value as it allows the process experts to do the following:

• Analyze and segment historical data quickly
• Perform complex realtime calculations on process data
• Improve knowledge sharing by allowing users to leave comments on process data (to contextualize data)
• Monitor and predict processes within the operational context
• Perform efficient root cause analysis
• Improve asset reliability

After an incident occurs, an incident investigation is usually conducted by operations to understand the root cause and generate action items that would mitigate or prevent future incidents. Traditionally, this has been done by manually scanning through process data in the historian and data wrangling in Excel files to establish a root cause or generate a correlation. Using advanced self-service analytics, root-cause analysis can be performed with significant time savings. With the click of a few buttons, it can search through historical data in a plant’s historian for patterns and specific process conditions, quickly locating all correlations between process tags (including lag time). Furthermore, the work can be shared between users, allowing for a bonus of improved collaboration.

Opportunities to improve safety

Advanced analytics offers a number of opportunities for improved safety, including the following:

Efficient process hazard studies. Process hazard analysis (PHA) studies are required by Occupational Safety and Health Administration’s (OSHA; Washington, D.C.) Process Safety Management (PSM) and the U.S. Environmental Protection Agency’s (EPA; Washington, D.C.) Risk Management Plan (RMP) regulations in the U.S. and serve as the foundation for process safety and risk management programs within chemical plants. PHAs also help protect against product quality issues, process downtime, and property damage. During a PHA session, the process experts often need to locate specific information or process parameters at a given time within the past 5 or 10 years to assess whether a particular process safety scenario is valid. Using advanced self-service analytics, the user can quickly pull up statistical information on and between different process tags and also layer events on top of each other, which facilitates the speed and increases effectiveness of PHAs. Another common task during PHAs is to identify how many times a certain temperature or pressure exceeded a given threshold. Traditionally, this could take hours if not longer to accomplish with multiple people looking at a span that could be 5 to 10 years. However, with advanced self-service analytics, this search can be completed in seconds. Furthermore, advanced analytics can help ensure that no event of interest is missed when assigning risk categories, which assists in the effectiveness and accuracy of PHAs.

Improved asset reliability. Picture a classic asset-reliability use case with two critical ethanol pumps in parallel, which deliver a high-pressure recycle stream to a packed-bed reactor (Figure 1). The primary indication of a leak is a decrease in flow, which sets off a distributed control system (DCS) alarm that alerts operators to a potential leak upstream or a pressure safety valve (PSV) release. However, the outlet flow meters occasionally drift, so operators are not always able to respond quickly enough to swap pumps before a large spill (that is, a process safety incident) occurs.

Figure 1. This photo shows two critical ethanol pumps in parallel that deliver a high-pressure recycle stream to a packed-bed reactor [6]

Using advanced self-service analytics, a realtime theoretical flow can be easily determined, and a deviation alarm can be created to alert production engineers and operators whenever the flowmeter needs to be recalibrated. Furthermore, it can be used to monitor the relationship between differential pressure and flow rate across the pump in real time, which serves as a secondary safeguard to ensure that the pump is operating on the pump curve. As a result, self-service analytics can help mitigate the severity of future process safety incidents related to these two ethanol pumps and also improve asset effectiveness.

Root-cause analysis for an ammonia PSV relief. Another common use case is performing root cause analysis to troubleshoot events that can lead to an ammonia PSV relief (Figure 2). Typically, production engineers would manually search through tens, potentially hundreds of process tags in the historian or DCS looking for a cause that led to a high-pressure relief event. This can take a considerable amount of valuable time and resources.

Figure 2. This piping and instrumentation (P&ID) diagram details the complexity and extent of troubleshooting required for an accurate root-cause analysis for an ammonia pressure safety valve [7]

With advanced self-service analytics, a single engineer or operator can quickly conduct a pattern recognition search on historical data to capture all instances of pressure transmitter reading spikes that led to the PSV relieving. Furthermore, advanced self-service analytics can analyze correlations between the pressure process tag and all process tags available in the historian, and account for time shifts between process tags. Once the process expert selects a process tag of interest as the root cause, it can be visualized and monitored in the future to minimize the potential for a PSV relief. The speed at which process safety root-cause analysis and incident investigations can now be performed using advanced self-service analytics is simply groundbreaking.

Concluding remarks

No one who works in or around a chemical plant ever wants to experience an incident. Safety is paramount in these plants and processes. Any tool or technology that can help facilitate and improve safety has great value in the CPI. Advanced self-service analytics can facilitate and improve process safety efforts, equipping manufacturing personnel with tools that result in process optimization, improved asset reliability, and even more effective process safety efforts. This is in line with what Huntsman, a leading global chemical company, has experienced. Advanced Analytics Manager and Global Excellence Team member for Huntsman Polyurethanes, Jasper Rutten, has stated, “… advanced analytics provides us with 24-hour engineering support, so we are able to optimize processes and asset reliability and to run our plants more stable… a more reliable and stabler site is a safer site.”

References

1. U.S. Chemical Safety and Hazard Investigation Board (CBS), Washington, D.C., www.csb.gov/williams-olefins-plant-explosion-and-fire-.

2. U.S. Chemical Safety and Hazard Investigation Board (CBS), Washington, D.C., www.csb.gov/chevron-refinery-fire.

3. The Oklahoman, www.oklahoman.com/article/3884430/fire-destroys-chemical-plant-in-thomas-ok.

4. New York Times, New York, N.Y., March 21, 2019, www.nytimes.com/2019/03/21/world/asia/china-explosion-jiangsu.html.

5. New York Times, New York, N.Y., June 21, 2019, www.nytimes.com/2019/06/21/us/philadelphia-oil-refinery-fire.html.

6. https://instrumentationtools.com/piping-instrumentation-diagram-pid.

7. https://fluidhandlingpro.com/canned-motor-pumps-in-use-at-petrochemical-plant-in-south-east-norway.

Author

Michael Chang is a data analytics engineer at TrendMiner (3701 Kirby Drive No. 740, Houston, TX 77098; Email: michael.chang@softwareag.com; Website: www.trendminer.com) He started his career in chemical manufacturing, with over six years of experience in operations, technology, process improvement, and environmental, health and safety (EHS) compliance. Having always been passionate about leveraging data to improve yield and troubleshoot manufacturing processes, he continues to drive improvements for the process industry. Chang holds a B.S.Ch.E. and a McCombs business foundation certificate from the University of Texas at Austin.