Engineering PhD Defense Questions: What Examiners Actually Ask

How did you validate your results — through simulation, physical experiment, or both — and how well do they agree?

This is a common opener for engineering defenses. Name exactly which results were validated against physical measurement and which were not, and state the quantitative agreement where you have it. If you have both computational and experimental data, say where they diverge and what you believe accounts for the gap. Claiming good agreement without citing specific metrics — residuals, percentage error, R-squared — invites an immediate follow-up.

You validated under laboratory conditions. How confident are you that the agreement holds under operating conditions closer to real deployment?

This is the external validity question for engineering work. Be precise about which parameters you controlled in the lab that would not be controlled in the field, and say what you expect that difference to do to your results. If you ran any sensitivity checks on those parameters, cite them. If you didn't, say what additional experiments would bound the uncertainty — don't claim the agreement transfers without evidence.

Your simulation produced results. How do you know the simulation itself is correct — as opposed to the model it implements?

This is the verification question: examiners expect candidates to distinguish between verification (did you solve the equations correctly?) and validation (are those the right equations for the physical system?). Describe your mesh or discretisation convergence check, your code verification against analytical solutions or benchmarks, and any manufactured-solution tests you ran. Treating the simulation as a black box that produces numbers is not sufficient.

What assumptions did you make in your model, and which of them are the least defensible for real-world conditions?

Name your three or four most consequential assumptions and give a one-sentence justification for each. Then identify the one that breaks first as conditions depart from your experimental scope. Examiners are not expecting perfect models; they are checking whether you understand the hierarchy of assumptions and can say what happens when the weakest one fails. Candidates who say their assumptions are all well-justified without ranking them lose credibility.

How sensitive are your conclusions to the boundary conditions you specified?

Describe the sensitivity analysis you ran — or, if you didn't run one formally, say what a range analysis would have involved and why you believe the effect is small. If you prescribed boundary conditions from literature values rather than measured them directly, name that as a source of epistemic uncertainty. Examiners in CFD, FEA, and similar fields treat boundary condition sensitivity as a standard check; arriving without having done it, or without knowing what it would show, is a gap.

How did you ensure the accuracy and calibration of your measurement equipment or test rig?

State your calibration procedure, traceability standard, and the calibration interval relative to your experimental campaign. If you used instruments that drift, say how you checked for drift during data collection. If a single instrument is responsible for the critical measurements in your thesis, the examiner will want to know its systematic error, its random error, and whether those propagate into your main conclusions in a way that changes the interpretation.

What testing or prototyping did you carry out, and what did it reveal that your simulations did not predict?

Answer this concretely — name a specific discrepancy, what caused it, and what you changed in response. Candidates who say simulation and experiment agreed perfectly on every point lose credibility; some discrepancy is expected, and recognising its source demonstrates understanding. If you used the prototype primarily to confirm simulation results rather than to discover new behaviour, say that and explain what the prototype added beyond confirmation.

Walk me through your uncertainty budget. What are the dominant sources of error in your measurements, and how did you quantify them?

A full uncertainty budget names Type A sources (statistical, from repeated measurement) and Type B sources (systematic, from calibration, assumptions, or instrument specifications) and combines them to an expanded uncertainty at a stated confidence level, following GUM practice or an equivalent standard. If you did not produce a formal budget, say what the dominant error sources were and give order-of-magnitude estimates. Saying results have 'some uncertainty' without quantification is not an acceptable answer at doctoral level.

How sensitive are your overall conclusions to a ±10% change in [a key input parameter]?

This probes whether your main claims are fragile or robust — in the genuine technical sense. State the sensitivity index or the direction and approximate magnitude of the effect. If your conclusion changes sign or loses statistical significance under a plausible parameter variation, you need to say that and explain what it means for the scope of your claim. If the conclusion is insensitive, say why — that is equally valuable information.

Could another research group reproduce your key results from the information in the thesis?

Reproducibility in engineering means specifying enough about materials, geometry, loading conditions, environmental controls, and data processing that an independent group could verify the result. Identify what they would need beyond the thesis — source code, fabrication tolerances, proprietary material data — and say whether any of those are available. Examiners are not expecting that every parameter was published; they are checking that you have thought about what is and is not recoverable from your written record.

Your error bars overlap between conditions X and Y. Can you still claim a meaningful difference?

This comes up whenever the quantitative margin between your result and a comparison condition is small relative to your measurement uncertainty. Respond with the statistical test you used, the effect size, and the engineering significance: a difference that is statistically marginal may still be practically meaningful if the absolute magnitude is large, or it may not be meaningful at all. Don't claim significance you cannot support, but don't concede a finding you can defend either.

What would a failure modes analysis of your prototype or system look like, and which failure mode concerns you most?

This question tests whether you have thought beyond the working case. Name the two or three most probable failure modes — whether from your own testing, from the literature on similar systems, or from first principles — and say what you know about the conditions that trigger each one. Examiners in mechanical, structural, and electrical engineering are particularly attentive to whether candidates have stress-tested their designs or only characterised them under nominal conditions.

What does your thesis contribute to engineering knowledge or practice that is genuinely new?

State one clear claim — a method, a result, a model, a design — and explain specifically why it was not derivable from existing work by straightforward application of known techniques. Naming a topic area is not a contribution; the contribution is what you discovered or demonstrated within that area. If your contribution is primarily empirical — you measured something that had not been measured — say what that measurement makes possible or changes in understanding.

A practising engineer argues that a simpler, cheaper approach already exists. How do you justify what you did?

This is among the most direct challenges an engineering examiner can raise, because it cuts to the heart of the research-versus-practice boundary. Name the simpler approach and say specifically what it cannot do in your context: it fails under the loading conditions you studied, it does not resolve the relevant length scale, or it requires instrumentation that is not feasible at the scales you worked at. If the simpler approach is genuinely adequate for most applications, be honest about that and explain what your work adds at the margins.

How does your work improve on existing methods — in efficiency, cost, reliability, accuracy, or performance — and by how much?

Give a number where you have one. 'Improved accuracy' is not a claim at doctoral level; '12% reduction in mean absolute error relative to the ISO 6336 standard method under condition X' is. If the improvement is not quantified, explain why — either the comparison is not commensurable or the relevant data for the comparison method was not available — and say what a fair comparison would require.

How did you handle the trade-offs between performance, cost, manufacturability, and feasibility in your design or method?

Examiners ask this to see whether you made deliberate decisions or simply optimised the metric that was easiest to measure. Name two or three explicit trade-offs you navigated and say what drove the resolution — a constraint from the application domain, a manufacturing limit, a cost ceiling, or a decision to prioritise the claim most relevant to the research question. Candidates who present their design as the optimal solution to all objectives at once are usually asked to identify what they gave up.

Is your contribution genuinely novel, or is it an incremental extension of existing work? Where exactly is the line?

Say where the line is yourself, before the examiner draws it for you. Name what the prior work achieved and name the specific gap your thesis closes. If your work is incremental — as most good engineering work is — defend why that increment matters: it enables a class of application that was previously out of reach, it resolves a known discrepancy in the literature, or it provides a validated tool where only unvalidated ones existed. Incremental contributions are not disqualifying; unacknowledged incrementalism is.

How could your work be adopted by industry, and what barriers would need to be overcome to reach commercial or operational scale?

Identify the gap between laboratory or numerical demonstration and deployment: certification requirements, material supply chains, manufacturing tolerances that your method demands, regulatory approval timelines, or the capital cost of the required equipment. Saying 'further work is needed' is not an answer. Name the specific barrier and say what form the further work would take. Examiners in applied engineering fields treat this question seriously and expect a candidate who has thought about the pathway from research to practice.

If you were starting this project today with what you now know, what would you do differently?

Give one specific, technical answer — a measurement you would have taken earlier, a model simplification you would have avoided, a parameter space you would have characterised first. The answer demonstrates reflective ownership of the research process. Candidates who say nothing would change appear defensive; candidates who list wholesale revisions appear to have lost confidence in the thesis. One concrete, early-stage decision with downstream consequences is the right scope.

Your model or prototype works at the scale you studied. What changes when you scale it up by an order of magnitude?

Scaling arguments matter in most engineering disciplines — thermal, structural, fluidic, electrical. Name the dominant physics at your study scale and identify which non-dimensional groups change unfavourably with scale: surface-area-to-volume ratios, Reynolds or Biot or Damköhler numbers, contact resistances, or structural loads proportional to volume while strength scales with cross-section. If you have not done a formal scaling analysis, say what one would need to include and whether you expect the current approach to hold, degrade, or fail entirely.