The snarled red lines on the chart look more like a plate of spaghetti than a source of fracturing insights.
It looks like a meaningless mess, which is generally how the ups and downs of difficult stages are viewed.
To Adam Hoffman, a completion engineer for Chesapeake Energy, those 47-stages-worth of data look like a valuable opportunity.
“We see so many stages with so many odd spikes and drops or chatter. We chop it off and say that was an odd stage. In my mind when we are looking at all those stages, we should wonder, ‘what was that pressure spike telling us,’” he said.
That curiosity became a research project after Chesapeake encountered a spate of blockages in recently fractured Eagle Ford wells. The investigation into the cause of the casing damage led to a collaboration with Well Data Labs to look for connections between pressure changes and what is happening in the wells.
Based on hundreds of stages of data from 19 wells fractured in the Eagle Ford, and later in the Powder River Basin, they reported finding a distinctive pressure signature that provides a reliable, but not foolproof, guide to when casing damage is likely.
Well Data Labs has automated the search for those signatures as it looks for the meaning of the terabytes of fracturing data in this overwhelming number of seemingly random, squiggly lines.
The oilfield data and software company is working on ways to monitor changes in the fracturing-fluid chemistry, the proppant intake into perforations, and an explanation for the pressure spikes seen before the pressure falls, said Jessica Iriarte, research manager at Well Data Labs.
The troubleshooting and pressure analysis were covered in paper SPE 201484 presented at the 2020 SPE Annual Technical Conference and Exhibition (ATCE).
It described how engineering troubleshooting revealed that geological stresses were the likely source of problems in one case, and faulty pipe in the other. It followed up with data analysis, which used machine learning to identify distinctive patterns that provide an early warning of what is happening in the well faster and more objectively than a completion engineer studying the chart.
Based on the troubleshooting, Chesapeake made changes that largely eliminated those costly problems.
But it was also a costly learning process. In the Eagle Ford, they identified the underlying problem by investigating why multiple coiled-tubing runs were blocked while they were trying to drill out plugs after fracturing. When that happens, Hoffman said, “it can mean a week lost working past it.”
Failure to drill out a plug can block access to the productive rock further down the lateral.
A reliable automated treating-pressure analysis in the daily report could alert the completion team to problems while fracturing is in progress. They could then make adjustments on later stages and create a plan to limit the time lost when drilling out plugs on stages where they are likely to encounter tight sections.
That will require an automated system that can spot the relatively few distinct pressure drops buried in thousands of stages. For the research project, that was done by manually scanning the data, which is an impractical option under normal circumstances.
What they have learned so far falls short of the industry’s holy grail—a real-time warning system that will allow those doing completions to make changes during fracturing before damage has occurred.
In early tests, the paper said the pressure signature is accurate in a high percentage of cases. Its shortcoming is that a potentially damaging pressure drop is instantaneous, said Mary Van Domelen, senior completions advisor for Well Data Labs.
The long-time completion engineer said the pressure drop happens in seconds. For those at the controls, she said, it is just long enough for the uttering of one, short expletive.
Working Backwards
Chesapeake began looking into casing damage when it ran into casing blockages in 11 of the first 67 wells in the Eagle Ford.
The problem wells were in acreage acquired from Wild Horse Resources in the Brazos Valley section of the play, which is east of the original development.
“We were learning as we were completing wells, and we quickly discovered abnormal tight spots as we were drilling out wells,” Hoffman said.
The evidence-gathering began by asking the drilling and geology teams if they had seen anything odd in those wells, he said.
The geology and geosteering data showed the stages with problems were in sections of the wellbore that came within a couple feet of the planes where formations in the lower Eagle Ford and Woodbine formation intersect.
This stood out because experience elsewhere had shown that bedding planes were prone to slip. Pressure pumping can trigger that damaging movement. They also noticed the rock had a high clay content, which is likely to swell while pressure pumping large volumes of water.
The problem stages were all located near the bedding plane, but not all stages near the bedding plane were a problem, so there were likely other variables.
To protect against bedding-plane stresses, they made changes to beef up the cement barrier, adding centralizers around the casing to ensure a sound, even cement layer. Clay-control chemicals were added to the fracturing fluid to limit swelling, and fracturing crews were directed to add crosslinked gels to the fluid when pressure increases were an issue.
At the time, Hoffman was using Well Data Labs’ application because it allowed him to overlay multiple layers of data to look for variables that moved in tandem during fracturing.
In the process of studying these charts, patterns in the treating pressure were identified in stages with tight spots.
“We noticed a big pressure build, a flat section, and rapid pressure drops” at a time when the pumping rate had not changed, Hoffman said.
Based on the detective work, they made changes. Drillers were told it was important to stay within the designed steering window and to avoid the formation tops. Completion plans categorized the risk of each stage and provided instructions for adjustments when problems were encountered during fracturing or drillouts.
The instructions included cutting a stage short. “What it came down to is, if they see anything funny at all, cutting that stage short rather than risking the entire wellbore,” Hoffman said.
The next 37 wells were completed without casing-restriction issues.
The second case study in the Powder River Basin began in much the same way as the first—there were six wells in which they were unable to completely drill out all the plugs using coiled tubing.
This time the questions quickly led to a prime suspect—bad-quality casing. Hoffman said the drilling crew mentioned that they had stopped using electric-resistance-welded casing due to manufacturing quality issues. They had switched to seamless tubulars, but there were a handful of wells among 11 that needed to be completed, so crews were told to treat them all similarly.
The pressure data again showed an instantaneous drop on stages with casing damage. Unlike the problems in the Eagle Ford, though, the overall pressure level was not consistently higher in the problem stages. Small splits along the seams of pipes were the likely source of the trouble, but they were unable to verify that with downhole imaging.
On the wells with the welded casing, the plans warned them of its risk of failure, and gave the crew options to change the fracturing plans if problems were encountered.
As in the Eagle Ford, the combination of investigation and plans that offered a map helped identify problems and eliminated those casing issues in future wells.
Abnormal Pressure Behavior
While the source of the trouble varied, they shared a common feature—abnormal pressure behavior.
The phrase was coined by Van Domelen, who offered an example while presenting the paper at ATCE: “The treating pressure rapidly increases at about 40 minutes into the stage, followed by an almost instantaneous rate drop.”
For Van Domelen and other completion engineers who have spent many hours in frac vans watching a line of pressure data slowly develop, that sharp pressure drop was familiar.
When those abrupt changes occurred, which was not often, they added to the worries during difficult stages where higher, volatile treating pressures demanded close attention.
Based on the work by Chesapeake and Well Data Labs, it is clear that not every sharp drop qualifies as an instantaneous signal of trouble.
Creating a program that recognizes the troubling subset of drops has required Well Data Labs to quantitatively answer the question: How fast does the pressure have to change to be instantaneous?
Using data from the field tests, plus other clients’ data since then, it measured the rate of change (psi/minute) associated with the stages in which obstructions had been encountered.
Some charts clearly show a pressure drop so rapid it looks almost vertical. But time series data from wells are often not so clear. The degree of change matters up to a point, but the rate of change was found to be more significant.
The model built by Well Data Labs removed noise in the surface readings. The paper acknowledges downhole data would be better, but the available surface readings are adequate.
To limit the noise in the data, a median value was plotted for the second-by-second data based on several seconds of data, similar to a rolling average. However, the outlier readings in the average were found to produce a less-accurate pressure line.
Measuring an abnormal pressure event required a model to precisely determine when the move starts and stops.
Well Data Labs is offering a model to identify abnormal pressure behavior after fracturing and is working on a real-time user application it expects to release early this year.
When the system was tested on 36 stages, it was right 70% of the time. It correctly identified 10 stages where the pressure signal accurately indicated that damage had occurred and 15 where it had not. Seven false negatives and four false positives were identified.
So far, this work is based on data from 12 wells and 486 stages, which were reviewed using a cloud-based application.
An optimist would say that more data will lead to a system that gets better at making these calls. A pessimist would say the sample was too small to know if that assumption is accurate.
What Is Normal?
After the presentation on identifying abnormal pressure behaviors, a participant in the virtual ATCE asked what normal pressure behavior looks like.
“You will get a different answer from every completion engineer,” Van Domelen said. To her, it is abnormal “if your pressure is behaving differently from other stages, even if you have plenty of room before reaching the maximum treating pressure.”
She and everyone else who spoke about this project are curious about the odd details noted in fracturing data.
“If you do have an abnormal stage that is an outlier, perhaps you should focus on it,” Hoffman said.
During the upcoming Virtual Hydraulic Fracturing Technology Conference and Exhibition, 2–4 February, Iriarte will be delivering a follow-up paper offering a look at Well Data Labs’ effort to translate fracturing data into useful indications of what is going on downhole.
This effort requires the speed and precision of computers and the guidance of experienced engineers with a deep understanding of the physical reality of fracturing.
When asked to describe what is happening as pressure rises and instantaneously falls, Van Domelen said it starts with pressure building because “a blockage in the well is developing. Pressure builds, pressure builds, pressure builds and … boom!”
The force, like a balloon popping, occurs in a complicated environment which shapes how that sudden release affects the well. In one of the paper’s case studies, the geology played a key role, while in the other, weak points in the casing were identified.
“With a lot of it, we do not really understand why it happens. It is very early on,” Hoffman said.
ConocoPhillips has also been looking for a pressure signal for damaging events downhole. Some are associated with rapid pressure drops, but that is not always the case.
“We have seen quite variable pressure and rate signatures. It’s not always a pressure drop or rate increase, as premised in this paper, but we have seen these types of signatures too,” said Eric Davis, global completion chief for ConocoPhillips.
He said the rock fabric has a significant impact on the pressure and rate response, and patterns can vary from field to field.
While the focus in the Chesapeake study was on pulling insights from the pressure data, it also showed that even with imperfect data, fracturing performance could be significantly improved with plans that offer more detail on the risk factors and how to respond to them.
“Everyone wants to talk real-time monitoring of jobs,” Van Domelen said. For that information to be useful they will need to know how to interpret the data and what to do. For the completion supervisor, she said, a critical question is: Do you have the authority to make decisions? Assuming you have the authority: If you see a problem during a stage, can you shut down, stop, and move to the next stage?
Seven Steps To Interpreting Fracturing Pressure Data
Mary Van Domelen, Well Data Labs
Editor’s Note: Mary Van Domelen, a senior completion advisor for Well Data Labs, compiled this guide to SPE technical papers about the use of fracturing data to interpret what is happening in the well. She recently coauthored and presented a technical paper which describes the characteristics of pressure signatures associated with events likely to cause casing damage.
SPE 201484 Identifying Casing Failures with Signature Fracturing Treatment Pressure Behavior by A. Hoffman, Chesapeake Energy; C. Wolfbrandt, J. Iriate, and M. Van Domelen, Well Data Labs.
The Mechanics of Design and Interpretation of Hydraulic Fracture Treatments
This paper represents the first reference to fracture treating pressure analysis. Published nearly 60 years ago, the basic concepts outlined in the paper have not changed. It provides concise, clear descriptions of processes for determining and evaluating fracture treating pressure trends.
SPE 1106-G (1959) The Mechanics of Design and Interpretation of Hydraulic Fracture Treatments by B.C. Crittendon, Mobil Oil.
The Importance of Slow Slip on Faults During Hydraulic Fracturing Stimulation of Shale Gas Reservoirs
The paper presents the mechanics of plane slippage in unconventional rock in an easy-to-understand format. The mathematics are supported by laboratory testing and field observations. The critical postulation of the paper is that the prediction of how pre-existing faults and fractures shear in response to hydraulic stimulation can help to optimize field operations and improve recovery.
SPE 155476 (2012) The Importance of Slow Slip on Faults During Hydraulic Fracturing Stimulation of Shale Gas Reservoirs by M.D. Zoback, A. Kohli, I. Das, and M.W. McClure, Standford University.
Impact of Cyclic Pressure Loading on Well Integrity in Multistage Hydraulic Fracturing
This paper focuses on two primary areas of potential well damage: erosion and cyclic pressure effects on pipe integrity and cement isolation. Casing loads are simulated with a mathematical model, and the impact of the loads is evaluated with various stress models. The critical postulation of the paper is that industry standard casing design practices need to consider the effect of dynamic loading on casing integrity.
URTeC 2902463 (2018) Impact of Cyclic Pressure Loading on Well Integrity in Multistage Hydraulic Fracturing by D. Barreda, M.P. Shahri, R. Wagner, and G. King, Apache Corp.
Leveraging Cloud-Based Analytics to Enhance Near-Real-Time Stage Management
This paper introduces a novel and highly effective approach in the field of hydraulic fracturing optimization and demonstrates that significant design improvements can be made by evaluating stage performance in real time as the well is being stimulated. The benefits of the technology are demonstrated through case histories, one of which is the first published example of using fracturing treatment data to not only identify the time of the casing failure, but also the depth, all within 30 minutes of shutdown.
SPE 197105 (2019) Leveraging Cloud-Based Analytics to Enhance Near-Real-Time Stage Managementby P. Bommer, Abraxas Petroleum Corp., and J. Iriarte, Well Data Labs.
The Unconventional Unconventionals: Tectonically Influenced Regions, Stress States, and Casing Failures
This paper provides an excellent overview of casing deformation related to hydraulic fracturing of horizontal wells completed in unconventional reservoirs. Topics covered include a review of the failure mechanisms, impacts on project economics, identification/quantification methods, and global case histories.
SPE 199710 (2020) The Unconventional Unconventionals: Tectonically Influenced Regions, Stress States, and Casing Failures by A. Casero and M. Rylance, BP.
One Stage Forward or Two Stages Back, What Are We Treating? Identification of Internal Casing Erosion During Hydraulic Fracturing—A Montney Case Study Using Ultrasonic and Fiber-Optic Diagnostics.
This is an excellent case study from the Montney which incorporates various diagnostic techniques to identify the cause and impact of casing failure mechanisms. Fracture treating pressure analysis is combined with DAS responses to relate changes in surface treating pressures to loss of confinement due to frac plug failures. Fracturing treatment pressure analysis proved to be an effective way to diagnose casing integrity loss events.
SPE 201734 (2020) One Stage Forward or Two Stages Back, What Are We Treating? Identification of Internal Casing Erosion During Hydraulic Fracturing—A Montney Case Study Using Ultrasonic and Fiber-Optic Diagnostics by M. White, K. Friehauf, D. Cramer, J. Constantine et al., ConocoPhillips.
Employing a Suite of Machine-Learning Algorithms in a Holistic Approach to Trouble-Stage Recognition and Failure Diagnostics
This paper, which will be presented at the 2021 SPE Virtual Hydraulic Fracturing Technical Conference, 2–4 February, explores a holistic approach to characterize trouble stages by applying automated event recognition of abnormal pressure increases and associating those events to formation and operational causes.
The paper will be available in OnePetro following its presentation at HFTC.
SPE 204145 (2021) Employing a Suite of Machine-Learning Algorithms in a Holistic Approach to Trouble-Stage Recognition and Failure Diagnostics by J. Iriarte, Well Data Labs; M. McConnell, Hawkwood Energy; S. Hoda, E. Siegel, and C. Wolfbrandt, Well Data Labs.
