Nine examples of effective process applications using Big Data

Last posting, in part one of this three-article series, we learned about Big Data — what it is, how it supports analyses, and how it can be helpful when we have to visually depict large volumes of data. In this article, we will look at nine actual situations from industry of how big data has been used in troubleshooting, maintenance and operations applications.

By Deborah Grubbe, PE, CEng.

Key to this investigation is the use of the Dynamic-Risk Analyzer (DRA) from Near-Miss Management, an advanced warning and risk detection software platform that identifies process problems at initiation stage.

TROUBLESHOOTING

Example 1: Plugging Equipment. The Dynamic Risk Analyzer was used to identify irregularities in fuel pressure to a burner over a period of several weeks. A shutdown was very expensive, so the operations group wanted to be certain of the issue to keep the shutdown as short as possible. The associated flow variable did not show any increase or upward trend during this time. The operations engineer noted that the DRA results could be an indication of plugging in the burner, but there was a level of uncertainty. So, in order to validate this observation, a new calculated variable was created in the analyzer to monitor the plugging in the burner.

In the following weeks, the operating team noted that the calculated variable exhibited several consecutive anomalies, thus confirming the plugging behaviour in the burner and providing valuable information. The operating team submitted a work request, and then planned and executed a brief shutdown to have the burner cleaned. The DRA data enabled verification of the data, which supported a brief shutdown that resolved the problem. Big Data analytical tools like the Dynamic Risk Analyzer allow other variables to be created to help support troubleshooting. The power of creating the variable, and then recreating the variable from past data, give the troubleshooting and maintenance teams an extra level of certainty that they may not have had.

Example 2: Cooling Water Loads. Anomalies were identified in product temperature and with associated valve movements for two consecutive days as the ambient air temperature increased.

Investigation pointed out that the product temperature was about 25 degrees higher than it was supposed to be, despite the coolant valve becoming maxed out (wide open). Both situations were not normal, and the product was not getting cooled properly despite the full flow of coolant.

This situation was caught right at the onset by the operating team using the DRA indicators. They concluded that because those two days were hotter than normal, the product didn’t get cooled properly. This was important to product quality. This was an important identification for the operating team, which enabled them to prepare to deal with the issue as weather continued to warm up for the summer.

Big data analytics aided the team in finding a potential issue early, even before the product was shipped. Their quick action enabled them to take other actions and to examine options to rectify the situation: e.g. larger valve, larger piping, and/or capacity of their cooling system.

MAINTENANCE

Example #3: Detecting Sensor Faults. Upon investigation, the team found an instrument fault with one of the sensors. They then informed the Instrument Group to take the necessary corrective actions to fix the faulty sensor and possibly address the factors that led to the fault. There happened to be three redundant sensors measuring this important level, and if 2 of the 3 sensors tripped, it would trigger the unit for an emergency shutdown. This meant that if two out of three sensors had become faulty or had fallen below a certain threshold, the ESD (emergency shutdown) system would have tripped the unit. Given the high likelihood of other sensors to get affected and become faulty too, the team recognized the importance of this early detection.

Example #4: Detecting Seal Oil Leakage. The operations engineer noted that although the variable had not crossed its alarm thresholds, the level was fluctuating significantly compared to other seal oil reservoirs. Upon investigation, the team found out that one part of the system had an oil leak. The operating team identified it as a critical issue and maintenance was able to fix the issue quickly.

Example #5: Avoiding Transmitter Failure. The operating team studied the DRA output from the previous day and were puzzled by the variability they were seeing in one of the control loops. After discussing it with both the maintenance and the operations teams, they realized that they had not made any purposeful changes that would cause this change. Upon investigation, these anomalies were found to be early indicators of the transmitter going bad. This situation was caught and addressed in a timely manner, with little to no interruptions to the operation.

OPERATIONS

Example 6: Equipment Operation. The DRA identified anomalies in overhead and reboiler temperatures in a column equipment set for several consecutive days. The operating team investigated the situation and found out that the product feed to the column contained a lower-boiling point impurity, causing the drop in temperature. This indicated dilution in the product concentration, which was undesirable for economic reasons.

To correct this problem, the operating team took action in the upstream unit, where the feed for the column was being processed. By adjusting the pressure in the upstream unit, they were able to reduce the impurity concentration in the feed. These corrections resulted in normal overhead and reboiler temperatures. Since the temperatures drops were very subtle and not easily identifiable by the naked eye, the economic savings would have been impossible without DRA.

Example #7: Catalyst Regeneration, Poor Shift Turnover. In reviewing the DRA data for an earlier shift, the operator identified a major upset in a catalyst regeneration column. Unfortunately, this upset had not been effectively covered at the shift turnover; however, the analyzer output had completely identified the issue. Upon investigation, the problem became clear: while they were feeding catalyst from a drum to the regeneration column (as part of the standard operating procedure), the feed rate dropped down to zero, yet the column kept running, making the process unstable. However, the instability of this procedure became clearer with the DRA indications. As a result of the data and further discussions, the operating team decided to modify the standard operating procedure to a more stable one. In the following months, no upsets were reported during the catalyst transfer. Also, some adjustments were made to what was discussed at the shift handoff meetings.

Example #8: Recycle Flows. The DRA identified anomalies in a recycle flow for two consecutive days. It turned out that there was a control loop issue where the flow was adjusted manually, and the associated valve movement was done automatically. The supervisors and operators were aware of this issue, but it had not brought it to the maintenance crew’s attention until the extent of the growing issue was identified and pointed out by the analyzer as a high-risk item. Following these indications, the operating team submitted a maintenance work order for correction.

Example #9: Unexpected Operator Action. A daily review of the analyzer output showed some abnormalities in a process unit. The unit appeared to have been running normally for the past few days, and these changes caught the curiosity of the engineer assigned to the unit. When the operations engineer tried to identify the cause of this issue, he noticed that one of the operators had, with all good intentions, moved the level setting from X% to Y%, although the operator wasn’t specifically instructed to do so. Luckily, this did not lead to a significant process problem; however, this did call to attention the need for some discussion around indicated issues with operator training and communications.

The DRA allowed the operations engineer to easily address a management system issue by highlighting this problem. This action led him to speak with the board operator, which led to greater understanding as to why this change was made without any instructions. It also led to a more detailed setting of expectations around how to communicate changes taken in the process, and what should and should not be indicated in the daily log.

As has been demonstrated by these examples, big data analytics can truly improve the fine detail in how your operation is managed. Big data analytics places all members of the operating plant on the same page and improves operations across all functions.

In Part III of the series, we will take a look at the leadership required to install a big data analytics tool.

Deborah Grubbe, PE, CEng. is owner and president of Operations and Safety Solutions (OPSS), a global consultancy that works with various industries. Grubbe is a former member of the NASA Aerospace Safety Advisory Panel and worked on the U. S. Chemical Weapons Stockpile Demilitarization. She serves on numerous advisory boards and is an Emeritus Member of the Center for Chemical Process Safety.