Unplanned downtime is a phrase that can strike fear in any upstream operator. A single unexpected pump failure can cost hundreds of thousands of dollars every day, not to mention the expense of emergency repairs. No matter where it occurs in the supply chain, unplanned downtime seriously impedes the efficient, dependable retrieval and distribution of petroleum products and services. So how can operators address such an expensive problem if they do not know how to see it coming?
The answer is Internet of Things (IoT) technology. Using machine learning and device sensors, IoT solutions can analyze the enormous datasets produced by assets in the field, model how those pieces of equipment work and interpret the myriad intertwined factors that influence their behavior. Operators can use this information in multiple ways to cut down unplanned downtime and improve productivity:
- Error prediction: Understanding the leading indicators of a failure event makes it possible to monitor for at-risk conditions as well as take proactive action to fix a problem or limit damage if the signs appear before a failure occurs;
- Error diagnosis: Failure prediction provides contextual data and models that are directly applicable to identifying the corresponding root causes as well as the steps and parts needed to make repairs faster, which enables quicker returns to production while lowering service costs; and
- Operation enhancement: These same data models also can help create optimal performance baselines for each asset. Comparing this to real-time operations makes it possible to identify potential problems and prescribe solutions more accurately, enhancing productivity.
Predicting equipment failures
The use of technology is not new in oil and gas, but historically much of the focus has been on hydrocarbon identification and extraction or managing individual assets. With the drop in oil prices over the past several years, upstream operators are working to improve recovery rates using methods and equipment that makes E&P more complicated and introduces more points of failure. Electric submersible pumps (ESPs) used for artificial lift are a perfect example. They allow increased flow rates for greater well productivity, but downhole failures result in expensive well intervention operations and lost production. It is critical for operators to understand how equipment is performing and the interactions that happen with it before a failure.
Data provided by connected ESPs and other business-critical equipment is the key to predicting failures, but the geysers of data coming from machines can quickly overwhelm the people trying to make sense of them. Workers tasked with analyzing equipment data can spend anywhere from two-thirds to three-fourths of their time on nonproductive tasks like simply trying to find the right data to analyze to find a problem. With that approach, it can often take weeks or months to get to the root cause of a single problem.
IoT can help by applying advanced analytics and machine learning techniques to this torrent of data to rapidly recognize patterns and anomalies. It can parse through endless streams of noisy, irrelevant data to detect small pattern deviations that may indicate changes in a machine’s state. Subject matter experts (SMEs) can then spend their time examining data more likely to be relevant to determine if the condition requires corrective action.
Over time, the system looks at the state of each individual machine from its first in-service date to its last along with error conditions and failures that occur during its lifetime. This process of mapping individual and groups of machine states and patterns and how they relate to each other enables the creation of digital models, or twins, of each machine. The models can be asked about the state of the machine and the probability of state changes that may foretell a failure. Time ratings can then be assigned to the state changes, effectively calculating a predictive failure scenario and giving the operator the ability to plan for repairs before the failure occurs.
Once an operator is able to get better insight into its equipment and is receiving alerts of impending failures, the next step is to pull in contextual data that can help determine exactly what is causing the problem and the best way to correct it. Awareness of a potential failure and performing repairs on time with the correct replacement parts on hand ensures continued production with minimal interruption.
IoT can tap into external data that add context to error codes, from environmental conditions (e.g., was there a period of extreme cold at the location?) to engine specifications and maintenance records (e.g., was a part recently replaced that might be bad?).
This added information narrows down the root cause of failure and helps identify the part or parts that need repair. Further linking to other information sources such as enterprise asset management systems can help locate parts quickly and automate inventory management. Detailed repair plans can be created that allow technicians to arrive onsite with the correct parts and step-by-step instructions that greatly improve the odds of a first-time fix.
Some equipment operates better than others. This could mean it lasts longer, has fewer failures, is more energy-efficient or has better output. Given a particular set of conditions, whether environmental, geothermal, aboveground/below ground or offshore/onshore, it is possible to accurately tune and calibrate equipment so that it is operating at its peak.
An IoT system can compare a group of like equipment to not only understand variations but also what drives those variations and provide prescriptive recommendations. For example, increasing pump rpms could remediate a drop in performance without any detrimental effects. The system can alert the operator and recommend parameter changes to increase those rpms.
As the operator reacts and makes decisions on actions, the system can learn to perform actions based on past responses and become more automated. This progressive approach allows operators to guide responses and understand cause and effect while ultimately becoming more proactive as the system becomes more intelligent. Using digital models, an optimized baseline is created that can be compared to equipment in the field to prescribe actions that elevate performance and increase productivity.
Partnering for success
Many operators are interested in improving their business with IoT but lack the skill sets and experience to be successful, especially since these concepts are not traditionally a core competency in this industry. It is important to understand that adopting IoT is a progression that involves readiness at the organizational level and the involvement of SMEs to guide the application of analytics, machine learning and modeling behavior as well as ties into other sources of enterprise information.
Advances in software can provide much of the heavy lifting of data analysis and free up operator experts to apply their knowledge to operationalizing the data back into the business. The most successful initiatives occur when operators partner with solution providers that can help analyze the vast amount of data available and then link to related areas of the business. This approach can generate significant, tangible business benefits, including optimized asset efficiency, valuable real-world feedback for R&D and substantially improved uptime, driving toward that holy grail of zero unplanned downtime.