Q1. Which of the following is a typical engineering data source used in AI models?

1. Randomly generated text paragraphs.
2. Time-series signals from accelerometers mounted on rotating machinery.
3. Only social media comments about engineering.
4. Hand-drawn sketches without any digitisation.

Select one:

A
B
C
D

Q2. Which statement best describes a physics-informed machine learning (PIML) approach?

1. A machine learning model trained purely on images, ignoring all physics.
2. A model that embeds governing equations or physical constraints directly into the learning process.
3. A process of replacing all simulations with random numbers.
4. A text-based model that predicts engineering laws from poetry.

Select one:

A
B
C
D

Q3. Which model is particularly appropriate for mesh-based structural data from finite element analysis (FEA)?

1. Graph neural networks (GNNs).
2. Naïve Bayes text classifier.
3. Simple linear regression with no spatial structure.
4. k-means clustering on raw node indices only.

Select one:

A
B
C
D

Q4. (Short answer) Briefly explain why non-stationarity is common in engineering sensor data (e.g., vibration, temperature) and why it matters for AI models.

Q5. Which maintenance paradigm is most closely associated with using AI to estimate Remaining Useful Life (RUL) and trigger actions before failure?

1. Reactive maintenance.
2. Preventive maintenance at fixed time intervals.
3. Predictive maintenance.
4. Ignoring maintenance until complete breakdown.

Select one:

A
B
C
D

Q6. Which sensor is most typically associated with vibration-based bearing fault diagnosis?

1. Accelerometer.
2. Thermocouple only.
3. Light sensor.
4. Rain gauge.

Select one:

A
B
C
D

Q7. Which model type is particularly suitable for learning temporal degradation patterns and predicting RUL from sequential sensor data?

1. Logistic regression with no time context.
2. LSTM or GRU networks.
3. k-means clustering on static features only.
4. Naïve Bayes classifier.

Select one:

A
B
C
D

Q8. (Short answer) Give one example of a time-domain feature and one example of a frequency-domain feature commonly used in vibration analysis for fault detection.

Q9. Which statement best describes a digital twin in engineering?

1. An unchanging CAD drawing stored on a local computer.
2. A high-fidelity virtual representation of a physical asset that is continuously updated with real-time data.
3. A purely textual description of a machine.
4. A random simulation unrelated to the real asset.

Select one:

A
B
C
D

Q10. Why are AI-based surrogates used for CFD or FEA simulations?

1. Because engineering simulations are conceptually impossible.
2. To approximate simulation outputs much faster, enabling design exploration and online prediction.
3. To avoid using any physical principles at all.
4. To generate random outputs unrelated to the simulation.

Select one:

A
B
C
D

Q11. In multi-objective engineering optimisation, which concept is used to represent solutions where no objective can be improved without worsening another?

1. Random search region.
2. Pareto front.
3. Single-point optimum only.
4. Gradient-free plateau.

Select one:

A
B
C
D

Q12. (Short answer) Briefly describe one advantage and one limitation of using AI surrogates instead of running full high-fidelity simulations for every design.

Q13. Which of the following is a key objective of a robotic inverse kinematics solver?

1. To compute a random joint configuration.
2. To find joint angles that achieve a desired end-effector pose.
3. To calculate energy consumption from a power plant.
4. To determine material fatigue life solely from static loads.

Select one:

A
B
C
D

Q14. Which controller explicitly optimises a cost function over a prediction horizon and can handle constraints on inputs and states?

1. Pure proportional-only controller.
2. On–off controller.
3. Model Predictive Control (MPC).
4. Open-loop bang-bang logic with no feedback.

Select one:

A
B
C
D

Q15. In reinforcement learning for robotics, which element represents the mapping from states to actions?

1. The environment dynamics.
2. The reward signal.
3. The policy (π).
4. The discount factor (γ).

Select one:

A
B
C
D

Q16. (Short answer) Give one reason why using pure RL policies directly on hardware without constraints can be risky for engineering systems.

Q17. Which of the following best describes AI governance in engineering?

1. Making AI as complex as possible.
2. Ensuring AI systems are developed, validated, deployed, and monitored under defined processes, standards, and accountability structures.
3. Removing all documentation to keep models secret.
4. Allowing AI models to run without any human oversight.

Select one:

A
B
C
D

Q18. Which concept is central to Industry 5.0 as discussed in the course?

1. Maximum automation with zero human involvement.
2. Human-centric, resilient, and sustainable integration of AI and robotics with engineering systems.
3. Exclusive focus on fossil-fuel-based systems.
4. Removal of all digital technologies from manufacturing.

Select one:

A
B
C
D

Q19. Which of the following is a legitimate safety concern for AI-enabled engineering systems?

1. A predictive model occasionally updating its documentation.
2. An AI-based inspection system systematically missing certain defect types in critical components.
3. A dashboard using a different colour scheme.
4. Engineers reviewing model predictions.

Select one:

A
B
C
D

Q20. (Short answer) Describe one way in which responsible AI practices can enhance trust and provide a competitive advantage for an engineering organisation.