DrillComputer

The DrillComputer is a software application for surveillance of ongoing and historic drilling operations for the purpose of decision support.

Challenges with conventional surveillance

Sparse physical mesurements

Physical quantities are mostly measured at the drill floor, while the interesting events happen near the drill bit.

Measurement errors

Each measured quantity contains a variety of systematic and non-systematic errors. As an example, if the hookload is measured by a strain gauge on the deadline, the hysteresis error can be several tons in magnitude and change sign depending on whether the drillstring is hoisted or lowered.

Derived parameters are only valid in steady conditions

The most interesting physical quantities are not measured directly, but derived from surface measurements by simple functional relationships. Examples of this include weight on bit WOB, rate of penetration ROP, mechanical specific energy MSE and friction forces (i.e the difference between up- and down-weights). These functional relationships do not take into account the dynamics of the drillstring. They can only be trusted in steady stress states.

The DrillComputer mitigates all these problems at once by using an ultrafast drillstring simulator. By running multiple parameter settings, the software predicts the most probable movement pattern of the drillstring at any time during an operation.

Current features

Precise ROP

The conventional estimate of ROP involves approximating ROP by the feeding speed at the surface. This estimate is accurate if we average over hours or minutes, but it is not a precise measure of the instantaneous downhole ROP. For instance, if we tag bottom with a block speed of 40 m/hour, it can take several minutes before the downhole ROP has increased to the level of the surface feeding speed. This is because the weight on the bit needs to build up before the full speed on bottom is achieved. In the figure below, we see a DrillComputr simulation of this situation. The calculated ROP is denoted ROP_CALC, while the surface feeding speed is denoted ROP. We recognize that ROP_CALC has the form 1 – exp(-t/T), where T is the halfing time that is dependent on the formation hardness divided by the drillstring stiffness. After about eight minutes we are drilling into a harder formation.  We see that ROP_CALC decreases momentarily until a new level of WOB is achieved.

In a real field situation, the formation hardness as a function of depth is not known, in contrast to the simulation below. However, it can be predicted by running the DrillComputer multiple times with different hardness parameters. Each run results in different estimates of the downhole WOB. The software predicts the hardness by choosing the WOB estimate that comply with a surface estimated WOB.

Moreover, in a real field case, the feeding speed is not a constant as it in the plot below. In fact the surface ROP varies much and it can also be negative while the bit is still on-bottom. The true ROP will therefore appear as a smoothed version of the feeding speed. However, normal smoothing of surface ROP will not achieve the same results. This is because smoothing reduces spatial resolution, while simulations do not.

Precise WOB and on-bottom detection

The conventional way of estimating WOB from surface data is to compute the difference between off- and on-bottom hookload. This is a fair estimate,  given that the hookload measurements are precise and that the drillstring is rotating, i. e. wall friction is neglectable.

Unfortunately, hookload measurements can have severe errors. For instance, when the hookload is derived from a strain gauge on the deadline, the hysteresis error can be several tons in magnitude and have a sign that is dependent on the direction of the drillstring movement. Surface WOB estimates are therefore only reliable when both the on- and off-bottom hookload measurements are noted when the drillstring is moving downwards.

The extensively tested pattern recognition module in DrillComputer ensures that surface WOB is used only when it is reliable. However, in the case when surface WOB is not reliable we can always rely on the simulated WOB.

In the figure below, we see that the conventional WOB, denoted WOBX, is always positive, despite the fact that the bit is moved on- and off-bottom several times. This can be seen from the bit depth (DBTM) curve. We see that the driller is picking up four times and each time he picks up about half a meter. For a drillstring with this stiffness, this should result in a WOB reduction of about six tons. Notice the counter intuitive behavior of WOBX when picking up. On each pick-up, WOBX actually increases a couple of tons, before it finally reduces but never goes to zero. This is clearly a pattern of hysteresis error. The calculated WOB, denoted WOB_CALC, seems to be more in line with our physical intuition. It reduces exactly when the bit is picked up and it is reduced about six tons as expected. Notice also that the software accurately detects when the bit is on-bottom (dark green) or off-bottom (light green). Finally, notice also the ROP_CALC curve is always smooth and zero when the bit is off-bottom.  By comparison, the feeding speed, denoted ROPS, is clearly not a good estimate of instantaneous downhole ROP.

Precise MSE

The conventional way for predicting MSE involves a simple functional relationship containing surface ROP, surface WOB, rotation speed and surface bit torque (often just surface torque). This MSE prediction has serious deficits in these cases:

  • There are large changes in block speed, such as when the bit is getting on- or off- bottom. This has a huge impact on MSE, since MSE is approximately inversely proportional to instantaneous ROP.
  • The well is deviated and the surface torque is many times larger than the bit torque.
  • The WOB is off because hookload measurements are erroneous or the off-bottom hookload is not reset properly.

 

Consequently, a variety of averaging methods have been proposed to reduce the impact of the above deficits. The problem with averaging is always loss of spatial resolution. The practical solution to this problem is that conventional MSE monitoring requires a trained analyst for tedious interpretation. The DrillComputer avoids all these deficits by deriving MSE from the hardness that is used to match the surface estimated WOB. This avoids all the problems with how to smooth data before calculations. The result is a smooth curve without loss of spatial resolution.

In the figure below, we have plotted the conventional MSE (MSES),  DrillComputer MSE, (MSE_CALC), downhole gamma ray (@D_GR) and density (@D_DEN). The data is collected from a well in the North Sea. Data is plotted on a depth scale with a range 500 meter. The conventional MSES has some similarities with the GAMMA and the DENSITY plots. The MSES plot is not averaged and we see a large number of outliers even though data is plotted on a log scale.

On the contrary, MSE_CALC, do not have the same amount of outliers even though no averaging is done. This is because MSE_CALC provides useful output even in the case when the drilling operations is not steady.

Smooth and high resolution MSE estimates can be used to more precisely identify drilling breaks, formation changes and drilling inefficiencies.

Along string inspection

The along string inspection mode lets the user inspect parameters that vary along the drillstring. This can be axial and torsional movements, stresses, normal forces, friction forces and stick-slip effects. Moreover, with respect to hole cleaning, it can be useful to use the along string inspection mode to see which parts of the drillstring lay on the low side and which parts are located on the high side. In the movies below, we see some examples of along string inspection usage.

In the example above we see an example where the drillstring is pulled out with an axial speed of 1 m/s. The initial pullout movement is accelerated too fast and axial stick-slip happens. This can be seen in the drillstring velocity (DS_VEL) and bit velocity (BIT_VEL) plots. Notice also that the high side indicator (HIGH_SIDE), demonstrates that the drillstring is hammering up and down in the build-up section.

In this example we see the development of stress forces and torques in the drillstring as the bit is tagging bottom. This is a J-well and we see that the location of the neutral point (location of zero tension) has moved far into the build-up section when full weight on bit has been developed.

Future features

 

We expect that the feature list will expand rapidly in the years to come. A solid foundation is made and we have also started on a variety of other features, such as prediction of hole cleaning factors, vibrations, tool wear and more. We encourage you to take contact if you want to be part of steering our development.