Safer and Lower Cost Well Logging
Formation evaluation logs in real-time or post-drilling, without the operational risks and at a significantly lower cost.
Data Driven Well Logging
QLog provides a suite of synthetic logs including shear, compressional, density and neutron. As a post-drilling or real-time service, this solution can help reduce the operational hazards of logging and eliminates the need for nuclear or acoustic tools in the borehole. With QLog, logging becomes a more cost effective and safer process.
Safer and Lower Cost Logging with Machine Learning
QLog provides a suite of synthetic logs including compressional and shear sonics, density, and neutron. QLog can be run for vertical, deviated or horizontal wells. Qualification tests with supermajors have shown QLog to provide the same accuracy as LWD tools in both deep-water and onshore wells. Additional benefits to the oil company include no nuclear or acoustic sources in the well; and savings up to 80% of conventional logging costs.
QLog can deliver results within the same accuracy of LWD tools for a fraction of traditional methods.
No boots on the Ground
QLog can be run as SaaS solution or from our offices in Houston, TX with no need for personnel in the field.
The data driven approach
employed by QLog allows for real-time near bit rock property predictions.
No Nuclear Sources
Since QLog's predictive power comes from a state-of-the-art neural network, we can record complete logs without having to send a single nuclear source downhole.
Drilling Data goes in, Rock Properties Come out
QLog is based on the simple principle that a fundamental physical relationship exists between drilling (EDR) data and the formation's elastic properties. Our approach, leverages neural network technology to uncover this relationship and derive triple-combo FE-logs from input EDR data. Key aspects of QLog have been awarded patent protection. This reinforces the growing importance of applying artificial intelligence to extract valuable rock properties information from the drilling data – and Quantico’s technology leadership in this area.
Conditioned drilling data is used as an input to a trained neural network. The results are high resolution formation properties.
Frequently asked questions
Where have you deployed your QLog technology?
In many of the unconventional plays in the US Land market. In addition, we have experience in Alaska, GOM, Offshore Canada, Oman, Saudi Arabia, Brazil, Australia, Malaysia, Norway.
If my logs are of poor quality, does QES have the ability to clean them up?
If competent EDR, Gamma and mud logs are available the answer is yes. QES is able to build logs from these data sets and provide infill logs where there are gaps.
How do you measure quality in your predictions of synthetic logs and calculate your errors?
Typically, we use Mean Squared Error or Normalized RMSE. But we can use whatever measure our customers are comfortable with, e.g., R2.
In many plays the Gamma Ray and Thrubit logs often simply mimic one another so how much additional information does the EDR data really add?
This really depends on the nature of the basin. If Gamma is highly correlated to Den/Son, then of course your statement would be true. If the attributes are not correlated, then we see results where combining GR with EDR clearly generates better results than either alone. We generally see that GR with EDR presents a higher quality prediction across all intervals, whereas GR alone may work in some but not in all lithologies.
Are there any data inputs that are different than typically used, e.g., ROP?
QLog uses any properties that are logged or measurements that are made in the wellbore. ROP, ROR, WOB, SPP, TOB, Temp, Caliper are all typically used as training data.
Is the Qlog data good enough to tie to rock physics models? Is there an example?
The DTC, DTS and RHOB simulations are often within a few percent of the measured data so if the measured data fit a rock physics model then so will the simulated data. However, it should be pointed out that since this is a completely data-driven process and if the trained dataset is sufficiently robust, the outputs may prove to be more reliable than a rock physics model.
© Quantico Energy Solutions 2020