Self-service analytics support asset reliability and operational performance
Case study examples demonstrate real world value
The oil & gas industry has seen, and overcome, some tough times in the past, but with the COVID-19 world economy and global drop in oil prices, it has taken a brutal beating.
There are ways to ease the burden — solutions that allow the use of data to reveal opportunities to reduce costs, waste and emissions, as well as improve operational and personnel efficiency and safety.
The key to that, however, is making sure the entire team of process experts and engineers not only have access to the data but also have the right tools to analyze it.
The goal of any oil refinery is to achieve operational excellence, day in and day out. That goal is within reach through use of a self-service industrial analytics platform incorporating artificial intelligence and machine learning to analyze plant data.
Data tells a story
A self-service advanced analytics solution may sound complex, but it doesn’t have to be. It’s made for users who need to get the most out of the data but are not necessarily trained to be data scientists. It uses pattern recognition along with machine learning to look at time-series data sets, allowing users to gain insights into what the data is saying. And these new insights drive action — from larger, data-backed decision-making to troubleshooting day-to-day occurrences.
Self-service industrial analytics empower those who work directly with operations to combine their expertise with the data available. It lets them gain knowledge from the data, to understand the story it is telling. But in order to get the full picture, data from a range of sources — time-series, contextual, operational — needs to be integrated, used and shared across all horizontal and vertical levels of the organization. Little can be gained from information sitting in silos. Like a mystery story, solving the crime depends on knowing all the facts.
When process experts easily gain insight into their processes using a self-service analytics tool, they can solve more day-to-day questions independently, without the help of data scientists. They can enhance their own effectiveness, saving time and money, and provide their organizations with new insights based on the ability to interpret what the data is telling them.
Maximized industrial manufacturing operations are achieved by more efficiently leveraging both human and data-rich resources. Value is delivered to owner-operators at all levels of the organization. To better understand this huge operational and profitability potential, we’ll look at three use cases demonstrating the application of a self-service industrial analytics software.
An unexpected flare flow problem
One day not long ago, an oil rig operator in the North Sea experienced an unexpected increase in flare flow at 10 am for about 15 minutes. Although not a critical event on its own, leaving the root cause unexamined could lead to more severe incidents in the future. Furthermore, more frequent occurrences of similar events could add up to significant losses.
In the past, the engineers would have turned to a more modeling-intensive approach to investigate the situation, but with self-service industrial analytics software, they were able to investigate and solve this problem with a couple of clicks. They highlighted the flare incident in the visualized data before them and searched for similar incidents. They found that 34 similar incidents had already occurred, which was a more frequent occurrence than had been realized.
Using the software, the team issued a report describing the investigative action taken and the results they found. They were also able to document the information for future reference and to alert co-workers on other shifts. Then the patterns of these incidents were overlaid to validate the similarity in process behavior.
Experts investigated the event further using the software’s search engine and feature calculations, helping them to better understand the severity and the maximum volume of the flare. As the software uses machine learning, it can hypothesize possible reasons for the issue. This helped the team to identify what was happening just before the flare and what happened when the flare occurred. They succeeded in finding the root cause of the event. Prior to any flare, a valve was closing too quickly. The experts easily eliminated the flare by changing the control logic.
The team had success because it proceeded methodically. It determined when such a spike has occurred before and the root cause of the incidents. It prevented similar incidents from happening in the future. By improving the critical valve’s reliability, they eliminated the source of the flaring.
Compressor trip diagnosis
Asset reliability and availability are key to meeting production targets, but this can be hindered by common production problems, like a compressor trip.
At a separation plant, total throughput decreased over the course of a month. Looking at plant-level key performance indicators (KPI), the central team and plant manager noted that one of the critical areas of the plant was having recurring problems: the sales gas compressors had been tripping more often in the last month. At the same time, during a recent shift, an operator reported a trip in one of the three gas compressors.
The team wanted to determine the root cause of the recent production losses due to the problems with the sales compressors, to enable automated proactive monitoring that could lead to greater uptime, and to enable process experts to analyze, improve and benchmark the process.
To find the root cause, a taskforce of plant process and reliability engineers launched an investigation using their self-service analytics software. They started the analysis by looking at the global key performance indicators (KPI) and the status of the compressor dashboards.
Next, they investigated the previous trip and found potential root causes using the software’s recommendation engine — a software feature that suggests possible root causes by analyzing historical time-series data of past events. Last, they used the multivariable pattern recognition feature to determine if the previous trips were brought on by the same root cause. They found that two out of the three trips in the previous month had the same cause — an increase in temperature after the fin-fan cooler. To solve this problem, they set a fingerprint monitor to provide an early warning for engineers and operators. And based on contextual information from maintenance systems and operators, automated recommendations for corrective actions were set as part of the proactive monitoring.
Improving sulfur recovery units
Sulfur recovery units (SRUs) reduce emissions and ensure compliance for operators. Encompassing a series of technical processes, they are becoming ever more important due to the rising demand for sulfur in various applications and the regulatory obligations around emissions control. Conventional means of monitoring the operational performance of SRUs and assessing the scale of any issues as they arise involve time-consuming, multidisciplinary data analysis projects.
One of the features of the self-service analytics tool is what is referred to as the “production cockpit,” which can be applied to monitor for when — for any of a variety of potential technical reasons — the non-converted hydrogen sulfide in the process is higher than expected, signaling a decrease in sulfur recovery rates.
In one instance, the process of finding the root causes associated with an underperforming SRU train — then making the necessary changes and introducing proactive monitoring — produced highly positive results, including 8% higher unit utilization, an increase in sulfur recovery from 96.5% to 99.3% and a 15% reduction in SO2 emissions.
In principle, this strategic approach of putting advanced analytics in the hands of subject matter experts allows handling of 80% of energy-related cases, contributing to corporate goals for reducing the carbon footprint.
With self-service analytics, it is possible to dramatically increase the number of data-driven decisions. With analytics capability closer to the process experts, for the majority of the cases no modeling is required to achieve a viable solution. Given current industry and market challenges, benefits also come from having a scalable and easy to replicate data analytics strategy to continuously improve the operational excellence.