In the past three decades, the so-called triple combo logging system—which measures formation openhole induction (resistivity), density porosity and neutron porosity—has become the standard logging tool of choice for reservoir engineers and logging analysts working in petroleum provinces all over the world. Inclusion of a sonic logging tool in a triple-combo string creates the familiar quad-combo suite of logs, which also has become an industry well-logging standard.

Yet, despite all the increasingly sophisticated logging capabilities at the disposal of the upstream oil and gas industry in the 21st Century, the ability to peer inside a well bore to measure such geophysical characteristics as formation porosity, deposition, dip, lithology, and fluid volume and properties can become a difficult task. In a very real sense, the ability to characterize a targeted geologic interval by interpreting downhole information depends upon the accuracy and coverage of the logging data, itself.

The fact remains that open-hole logging conditions encountered in some wells are so difficult that logging openhole via triple-combo or quad-combo technology doesn't always provide accurate information and in some cases misses important geologic intervals entirely. This is especially true in deviated wells, where washouts or tool sticking can interfere with obtaining the quality of data needed to perform an evaluation. In some extreme situations, poor downhole conditions can even result in the loss of an entire openhole logging run.

In addition to the technical challenges, there are also economic barriers to acquiring high-quality openhole logs in some wells, such as drilling programs with multiple casing strings in which openhole logging data must be obtained after each well segment is drilled. In multi-well drilling programs, the costs of difficult logging conditions—or well-construction limitations imposed by challenging or hazardous drilling conditions—can be magnified across an entire field.

Click image to enlarge

HES introduces Chi Modeling
In early 2002, the Logging and Perforating product-service line of Halliburton Energy Services commercially introduced a new post-processing service in the Rockies called Chi Modeling that helps producers better and more economically evaluate a reservoir despite poor borehole conditions or erroneous or missing data.

In essence, Chi Modeling is an extremely sophisticated, artificial intelligence technology in which proprietary algorithms are applied within a neural network to predict needed triple-combo or even quad-combo openhole data with a very high degree of accuracy. Input data for the Chi Modeling workflow consists of known triple-combo or quad-combo openhole data from a so-called training well and pulsed neutron data from both the training well and a nearby application well.

Chi Modeling begins reconstructing missing openhole data in an application well by associating openhole triple-combo data and cased-hole pulsed neutron data in a training well. A root mean square error statistical analysis is performed using processing weights assigned to each log curve in the training well to find consistent associations between the openhole and cased-hole data. The predicted values obtained are tested against the original openhole data in the training well. If the predicted values do not match the actual values in the reference well adequately, the weights are changed and the model is re-computed until a match is obtained.

The weighted relationships between each input variable are summed to predict openhole triple-combo or quad-combo data in the application well, for which only pulsed neutron cased-hole data is available.

This figure shows a comparison between the original neutron/density porosity data (Track 3) and the predicted neutron/density porosity data for a reference well Track 4, as well as the original black and predicted (red) 90-inch resistivity data (Track 2). Track 1 is the openhole gamma ray.

Analysts can use the weighted associations between openhole and cased-hole data from the training well to fill in data gaps in the application well where the original wireline or logging-while-drilling data is missing; to replace poor quality data resulting from poor borehole conditions; and to generate reliable openhole logs when none are available. The associations may be used confidently as long as the formation geology of application wells remains similar to that of the training well and the formation's geology is adequately sampled and represented in the training reference well. If formation geology in the application well changes enough, a new set of openhole data is required to create a new set of associations.

Applications of Chi Modeling
Since the introduction of Chi Modeling, producers have used the service in hundreds of applications to control costs and manage risks in a wide range of field development and redevelopment scenarios. To encourage operators to try Chi Modeling, Halliburton in many instances has proven its effectiveness with "blind" demonstrations.

Two wells are selected for each blind test, each with openhole and cased-hole logging data. One well is selected as the training well and Halliburton is given all the logging data from that well; the openhole data from the second well, called the test well, is withheld. Halliburton then uses Chi Modeling to associate openhole and cased-hole data from the training well to create a set of pseudo openhole logging data for the test well. Next, the pseudo openhole logging data created for the test well is compared with its actual openhole data set. Without exception, the pseudo openhole data generated using Chi Modeling in the blind test has matched the actual openhole data with an extremely high degree of accuracy.

In mature fields, Chi Modeling has been used to screen old wells to determine whether they are candidates for workover or should be plugged and abandoned. Evaluations of existing wells can be accomplished more quickly and usually at lower cost because an operator needn't pull tubing from an old well to log openhole. Instead, he can obtain the data he needs by logging through casing with the pulsed neutron tool and using Chi Modeling to convert the information to pseudo openhole log data.

The process also has proven useful evaluating new wells, especially in multi-well development programs where operators in some cases have sped the pace of drilling and avoided costly wireline openhole logging runs. An operator can perform a complete openhole formation evaluation in one well; install casing in the well and log using a pulsed neutron tool; then obtain only cased-hole pulsed-neutron data in other wells included in the development program and use Chi Modeling to create pseudo openhole logs across the entire field.

In one notable application of Chi Modeling, an operator had been unable to obtain openhole logs in the bottom 300-400 feet of wells throughout a whole field, including the targeted pay zones. First, the operator tried to solve the problem by using open hole logging tools to obtain as much data as possible in the well; then casing the remainder of each well and logging the remainder of the well bore with cased-hole logging tools. However, the openhole and cased-hole data couldn't be correlated with enough accuracy to meet his technical and economic objectives. Halliburton was able to create a pseudo open-hole log over the lower portion of the well using Chi Modeling based upon the cased-hole logs which, according to the client, was accurate to within one porosity unit across the pay sand.

Chi Modeling can also been used in earth-modeling applications to monitor evolving downhole conditions in primary-, secondary-, and tertiary-recovery projects.

Looking to the future
There are essentially no geophysical parameters that prevent Chi Modeling from being used in any downhole setting. The breakthrough technology uses familiar input data and generates output values that are easily recognized and understood by users.

Halliburton's Logging and Perforating PSL began introducing Chi Modeling internationally about five months ago, and applications are expanding around the world. The technology is at its best post-processing logging data for field redevelopments or new-field developments involving multiple wells.

Although Chi Modeling is used today somewhat after the necessary input data is collected in the field, future applications could occur closer to or even at the well site as reservoir engineers and log analysts become more familiar with the workflow and more adept at modeling associations between openhole and cased-hole logging data.