Gains:
- Ability to understand the responsibility of verifying AI output in the context of oil and gas legislation (EMRA, PRMS/SEC reserve reporting, API standards)
- Ability to transform commercial confidentiality, well data sensitivity and ethical boundaries into operational confidentiality rules
- Ability to sustain the use of AI in oil and gas engineering with an end-to-end verification and governance discipline
Module Exam
1. Which of the following is the fundamental principle that determines the risk of an AI output in oil and gas engineering?
- A) The risk of an output is equal to the harm it would cause if that output were faulty; validation scales accordingly ✔
- B) AI output is generally safe as it is written fluently and confidently
- C) Validation is unnecessary when the most current model is used
- D) If the output refers to a standard, no additional control is required
Explanation: The risk of an output is equal to the harm it will cause if that output is faulty. While a minor error in a literature summary may be harmless, an error in a well pressure limit, leak alarm threshold, or reserve report can lead to an explosion, environmental disaster, or major financial loss; so verification intensity scales with potential harm.
2. What does the problem of 'non-uniqueness' in seismic inversion necessitate the use of AI?
- A) Direct acceptance of the first interpretation produced by the model
- B) Narrowing the interpretation by connecting it to the well log, well tie and geological constraints ✔
- C) Select only the highest resolution slice.
- D) Complete abandonment of seismic data
Explanation: The same seismic reflection data can be explained by more than one different subsurface model (impedance, lithology, fluid distribution). So an interpretation produced by AI is not the only correct one; The well log must be narrowed down by connecting it with independent constraints such as well tie and geological context.
3. Which limit should be specifically kept in mind when estimating permeability (permeability) from the well log using machine learning?
- A) The model is automatically valid in every field and every lithology
- B) The relationship learned is specific to the educational conditions; ✔ Extrapolation to different sites carries risks and must be linked to core/well testing ✔
- C) Direct and error-free measurement of permeability from logs
- D) Forecast uncertainty should not be reported at all.
Explanation: Permeability is not measured directly from logs; The relationship learned by the model is specific to the facies and pressure conditions of the training data. Blindly applying (extrapolation) to a different site or unfamiliar lithology can cause major error; The output must be linked to core measurement and well testing.
4. What happens if the 'b' exponent of the Arps hyperbolic model is kept too high (e.g. b>1) and extrapolated for a long time in drawdown curve analysis (DCA)?
- A) EUR becomes lower than it is and the reserve is hidden
- B) The decline becomes unrealistically slow and the EUR becomes overly high, optimistic ✔
- C) Production data is automatically cleared
- D) Well pressure increases physically
Explanation: In the Arps hyperbolic model, the prime b determines how much the rate of decline slows down over time. Taking b>1 and extrapolating it to the long term will cause the decline to be unrealistically slow and the EUR (final estimated production) to be excessively high. Therefore, in the late period, a transition to exponential decline or a minimum drawdown limit is applied and the result is tested with the material balance.
5. Why is the Archie equation for porosity and water saturation the validation anchor of an AI output in petrophysical interpretation?
- A) Because it only shows the well cost
- B) Because it establishes the physical relationship between resistance, porosity and water saturation and tests the output ✔
- C) Just because it defines the color of the log
- D) Because it is only meaningful in a laboratory setting
Description: The Archi equation establishes the physical relationship between log resistivity and water saturation. A water saturation value suggested by the AI is physically tested by inserting it into the Archi equation with porosity and resistivity inputs and comparing it with local formation coefficients; The result should be in the range 0-1 and consistent with the core.
6. Why is a 'false negative' (miss) particularly critical in pipeline leak detection?
- A) Because it classifies the actual leak as 'none' and delays the intervention and causes a risk to the environment/life ✔
- B) Because it only changes the color of the SCADA screen
- C) Just because it increases the number of pages of the report
- D) Because false negatives are always harmless
Explanation: A false negative is a classification of 'absent' for a leak that actually exists; In this case, intervention is delayed and there is a risk of environmental disaster, fire/explosion and loss of life. A false positive results in unnecessary stopping costs. In safety-critical pipelines, the threshold is chosen cautiously to minimize false negatives and is supported by field confirmation.
7. In predictive maintenance, what is the most appropriate use when producing a 'remaining useful life' (RUL) estimate from vibration data of a compressor?
- A) Accepting an estimated exact failure date and not making any checks until that day.
- B) Taking the estimated uncertainty range as decision support and confirming critical equipment with field inspection and root cause analysis ✔
- C) Disabling the alarm completely when vibration increases
- D) If RUL is low, operate the equipment to its fullest without any uncertainty.
Description: RUL (Remaining Useful Life) estimate is a probabilistic estimate, not an exact date. Proper use is to take the estimate with its uncertainty range and use it as decision support in maintenance planning; In critical equipment, it is confirmed by field inspection and vibration spectrum root cause analysis. Continuing to operate equipment based on a blind date is risky.
8. Which principle is essential if AI-based gas detection is used in a site with a risk of toxic gases such as H2S (hydrogen sulfide)?
- A) AI detection could be the only layer of security, replacing certified physical detectors
- B) AI is an auxiliary layer; No substitute for independently certified detectors and procedures, multi-layered security essential ✔
- C) If the model confidence score is high, physical detectors can be turned off
- D) H2S detection can only be done with office analytics
Description: H2S is a deadly gas in terms of life safety; AI detection can be a helpful layer, but it cannot replace standalone, certified physical gas detectors and evacuation procedures. Security-critical functions should not be based on a single artificial intelligence model, but on multi-layered and independent security systems (defense in depth).
9. What is the main purpose when processing a free text drilling log report ('ROP dropped throughout shift, torque fluctuated, possible clay swell') with AI?
- A) Randomly rewriting the report
- B) Automatically increase drilling depth
- C) Making data queryable by translating inconsistent free text into standard fields; leave the comment to the engineer with the tag 'possible' ✔
- D) Physically renumbering the well
Description: AI transforms inconsistent free-text operational reports into standard fields (incident type, depth, indicator change, probable cause, action), making data queryable and comparable. However, the root cause and operational decision are confirmed by the drilling engineer; the model should retain the 'possible' label.
10. What is the correct approach when entering site-specific well log, production and reserve data into a general AI tool?
- A) Entering the actual coordinates and production values exactly and getting the fastest response
- B) Anonymizing identities and values, working with representative values and using only approved/controlled tools ✔
- C) Loading the entire field database and allowing the model to learn
- D) Note that the data is confidential and share it as is.
Description: Well coordinates, production flows and reserve information are commercially extremely sensitive and often confidential data belonging to the license holder/partnership; It is subject to public regulation in some countries. Entering an uncontrolled external service creates competition, breach of contract and legal risk. The correct approach is to anonymize identity/coordinate/actual values or work with representative values and use only approved, institution-controlled tools.
11. What does 'order of magnitude control' mean in an engineering calculation with AI?
- A) Increasing the number of decimal places of the result
- B) Testing with a quick hand calculation whether the result is roughly within the expected power of 10 range ✔
- C) Measuring the graphic resolution of the result
- D) Standardizing the color of the result
Description: Order of magnitude checking is testing by quick hand calculation whether the result is roughly within the expected power of 10 range. For example, the permeability of a sand reservoir is expected to be in the millidarcy range, but the occurrence of thousands of darcys instead of darcys reveals a unit or editing error without going into detail.
12. What is the most robust way to verify the result when printing production data analysis code to AI with Python?
- A) If the code works and does not give an error, accept the result as correct.
- B) Comparing with unit testing, physical limit checking and independent method with known small sample ✔
- C) Relying on the large number of lines of code
- D) Checking whether the graphic output is colored
Explanation: Although the generated code looks smooth and correct, it may contain hidden errors (wrong unit, shifted date, leaked data, wrong collection). The most robust verification is to manually calculate the expected output on a small known sample, verify it in the code (unit testing), check the intermediate results against physical limits (such as no negative flow rate), and compare them with an independent method.
13. What is the most accurate statement for using an AI-generated EUR forecast in reserve reporting (e.g. PRMS/SEC)?
- A) Assessor approval is unnecessary as the AI output is automatically valid
- B) AI is helpful; method, input and uncertainty must be traceable, documented and approved by the competent evaluator ✔
- C) The model with the highest EUR value should always be preferred.
- D) Reserve declaration can be completely automated and human control can be removed
Explanation: Reserve reports are auditable documents that are the basis for investor and regulatory decisions and are under the responsibility of the qualified reserves evaluator. AI can help with data cleaning, dream fitting, and scenario generation; however, the method, inputs, and uncertainty of the figures must be documented in a traceable manner and approved by the authorized evaluator. AI alone cannot declare reserves.
14. Which statement is most accurate regarding the ultimate responsibility for the use of AI in oil and gas engineering?
- A) Since AI produces the output, the responsibility passes to the software company
- B) AI provides support and acceleration; Responsibility for decision and report remains with authorized engineer, output must be verifiable and human approved ✔
- C) Engineer approval is unnecessary as the AI output is automatically valid
- D) Responsibility lies only with the field team, it does not concern the office engineer
Description: AI is a powerful tool in support work such as data processing, drafting, discovery and optimization acceleration; However, the responsibility for the signed report, safety-critical decision, well design and reserve declaration remains with the authorized engineer/evaluator. Outputs must be auditable, verifiable and human-approved.