EXPERT SYSTEMS HELP DESIGN CEMENTING AND ACIDIZING OHS

D.D. Onan, D. Kulakofsky, M.S. Van Domelen, W.G.F. Ford
Halliburton Services
Duncan, Okla.

Knowledge-based expert information systems can help train less-experienced designers and orient seasoned designers at new locations. These systems are playing an increased role in completion and production operations.

For Halliburton Services this computer software is at work in numerous areas aiding in the design of cement jobs and stimulation treatments. 1 2

Both seasoned operators and engineers in training call on expert systems to provide answers to complex, multifaceted treatment designs.

Expert systems help:

Design treatments based on an accumulation of knowledge from experts
Provide technical information and guidelines on the proper use of additives
Serve as a training tool for less-experienced personnel.

These systems incorporate the history and knowledge of a well and formation with state-of-the-art techniques, products, and recommendations.

Knowledge-based information systems rely on large accumulations of information stored and retrieved through rules and principles that match the information to such applications as cement slurry design and matrix acidizing.

Rules determine how to use the collected data to choose the products and processes required for a given well.

The software offers multiple selections for each category of material, while eliminating any incompatible additives from the category.2 In cases where products are not readily available on location, second and third options with all associated information may help satisfy operational requirements that benefit both the operating and service companies.

EXPERT SYSTEMS DESIGN

The process of creating a knowledge-based expert system for the oil industry is complex and exhaustive. But the benefits of obtaining immediate conclusive information warrants the time and expense.

Typically, a team of knowledge-acquisition experts, programmers, and designers meet to establish goals and requirements.

Some criteria for the cementing expert system included::z

Ensuring consistent designs by having one program worldwide
Adding new products easily
Operating in a user-friendly mode
Calculating slurry designs
Evaluating a wide variety of potential problems unique to a cement slurry design.

The process requires both data acquisition and system design. Data acquisition from the domain experts included team meetings, questionnaires, and face-to-face interviews.

Information was collected in 15 categories, as follows: 2

Cements
Salt
Chemical resistance
Lightweight cements
Fallback
Lost circulation
Gas migration
Expansion
Strength retrogression
Fluid loss
Accelerators
Retarders
Heavyweight cements
Dispersants
Defomers.

For the matrix acidizing system, three expert teams and one team of field advisors were organized. The domain-expert teams included a coordinator and four to six specialists with an aver-age of 12 years' experience.

Team members had degrees in either engineering or chemistry. Nine members had BS degrees, one had an MS degree, and seven had PhD degrees.

Each field advisor had a minimum of 10 years' diversified experience. Their experience represented knowledge of the Midcontinent U.S., Gulf of Mexico, California, Asia-Pacific, Europe, and Africa. These teams ensured that the computer program would be based upon accurate and diversified knowledge.

The three expert teams were responsible for: 3 0 Base fluid selection 9 Additive selection 0 Product data and information.

The knowledge-engineer had to first verify, organize, and convert the information to rules before the rules could be programmed into the computer system. Table 1 shows a sampling of the strength retrogression rules.

It was equally important to provide clear and thorough documentation to verify system behavior and to aid in integrating new products and processes.

After the team created a set of rules, a PC-based artificial intelligence (AI) shell was used to build the knowledge base and implement the rules.

The cement design software uses almost 1,000 rules to evaluate conditions and provide actions to induce changes and inferences in the overall system.

The system was designed after accumulating and coding the knowledge base. These knowledge-based expert systems operate with Microsoft Corp.'s Windows. A graphical user interface (GUI) simplifies data input with a point and click method.

The software is used either as a system-dictated recommendation or as a selective process using operator preferences. The conventional method is to enter the basic data and allow the system to recommend the optimum materials.

This system is especially helpful in unfamiliar environments, new situations, or for training. However, the feature that sets this software apart is the ability to customize the recommendations.

The user selection mode builds a list of available materials for the given conditions and presents them in groups on the selection screens. The user can then pick the desired product based on individual preference, customer likes and dislikes, and local availability.

Subsequent product selections are tailored around the previous decisions."

PRACTICAL APPLICATIONS

The software is designed to be updated easily. To allow consistent, technically sound, advanced designs at any location, new products and processes will be added periodically by the system programmers and product researchers. If the user questions why a certain product was recommended, a panel containing product data and tutorial information can be displayed.

The information can be read interactively, printed immediately, or stored to an output data file and printed with the final recommendation. Because the design can be restarted without re-entering input, the user can experiment as many times as necessary to evaluate several materials before ordering laboratory tests.

The program runs in the Windows PC environment with a GUI; therefore, no complex typed commands or codes are needed. A mouse can control the cursor to select buttons or on-screen menus that perform computer commands.

The user-friendly program helps ensure that the time spent in the program is to design the job rather the learning to use the program.

CEMENT JOB

A job performed on deep, hot well in South Tex as illustrates the interaction of the system, the professional engineer, and the support laboratory.

Because in higher temperature wells, above 400 F., each cement job is unique, extensive laboratory testing is often required to determine retarder response, fluid-loss control, thermal thinning, and settling of cement slurries.

Cement slurries must not only deal with these damaging effects, but also must maintain good Theological profiles to be effectively placed. Special additives are required. The system recommendations are a basis for determining further testing and fine tuning by the field laboratory.

Experience has proven that surface slurry properties and standard laboratory testing procedures are inadequate indicators of downhole dynamics. The resulting cementing composition can generally leave those unfamiliar to the practice perplexed at the wide array of additives contained in the final design.

To reduce laboratory time and effort in customizing a higher temperature job for a South Texas well with a 425 F. BHCT, the cement expert system was used to recommend proven methods for solving solids settling problems at the slightly lower BHCT of 400 F. (Table 2).

Because of the higher BHCT, the increase in retarder concentration, above 400' F., required the Standard one-to-one addition of a retarder intensifier and water.

The additional water content required:

More weighting material to maintain slurry density.
More fluid-loss control material to maintain minimum filtrate loss.
More high-temperature suspending aid and more settling control agent to balance optimum surface mixing viscosity with the desired downhole properties.

ACIDIZING

Several basic problems can reduce or halt production of wells. In old wells, declining reservoir pressures cause changes in reservoir forces and conditions that release fine particles near the well bore (plugging).

In new wells and worked over wells, formation damage may be caused by fluid loss (invasion) into the producing formation. Occasionally, other unexpected complications may reduce or block productivity.

Often acid Stimulation treatment experience is limited. The knowledge-based expert system can be used to match well history/damage evaluation, current well information, and proven methods and products for acid stimulation.

In a field in Kalimantan, Indonesia, a well had historically produced 300 b/d of clean oil. During a workover, unexpected fluid losses caused total loss of oil production.

The acid expert system matched well history, formation characteristics, and current information to recommend an acid stimulation treatment that was run with slight modifications to meet customer specifications and the engineer's professional judgment (Table 3).

Modifications included the following:

The customer's specifications did not include iron control. But the software recommended adding iron control additives because of siderite (FeCO3) in the formation. The actual volume was a compromise between the recommended and the customer's desire.
Other additives not previously included were a mutual solvent and clay stabilizer. These were recommended by the system and confirmed by laboratory testing. The engineer, laboratory, and customer agreed upon a volume.

After the acid stimulation, the production increased to 132 bo/d.

REFERENCES

Kulakofsky, D., and Crook, R.J., "Knowledge-based Expert System Eases Cement Slurry Design," Production Technology, June 1992, pp. 43 and 48.
Kulakofsky, D . S., Wu, F.L., and Onan, D.D., "Development and Application of a Knowledge-Based Expert System for Cement Slurry Design," Paper No. SPE 4417, SPE Petroleum Computer Conference, Houston, July 19-22, 1992.
Van Domelen, M.S., Ford, W.G.F., and Chiu, T.J., "An Expert System for Matrix Acidizing Treatment Design," Paper No. SPE 24779, SPE Annual Technical Conference and Exhibition, Washington D.C., Oct. 4-7, 1992.
Chiu, T.J., Caudell, E.A., and Wu, F.L., "Development of Expert Systems to Assist with Complex Fluid Designs," Paper No. SPE 24416, SPE Petroleum Computer Conference, Houston, July 19-22, 1992.