Mechanical Engineer Interview Questions for Experienced Candidates

“For the last three years I've owned the mechanical design of a packaging machine line, about four hundred parts, with two designers working from my layouts. Two decisions were mine. First, I moved the main frame from one big machined plate to a welded tube frame built in bolted sections, which cut lead time and made the frame easier to modify when customers asked for longer infeed sections. Second, I chose to standardise every linear axis on one guide rail size, even where a smaller rail would have done, so maintenance teams carry one spare instead of four. The line has run at customer sites for two years with no frame or guide failures. If I did it again, I'd spend more time on the guarding design early, because we reworked it twice after safety reviews.”

“Our programme manager wanted to drop the corrosion test on a new outdoor enclosure to hit a launch date, since the paint system was the same one we'd used before. What had changed was the steel supplier and a new folded seam design, and seams are exactly where corrosion starts. I didn't just say no. I showed him two warranty returns from an older product that had rusted from a seam, and what those cost us in replacements. Then I offered a plan: run a shorter corrosion test on the seam alone in parallel with launch preparation, and hold the first shipment only if it failed. He agreed. The seam showed rust early, so we added a sealant step before painting. We launched one week late instead of shipping a problem.”

“I estimated eight weeks to design and release a new clamping fixture, and it took fourteen. My design work was roughly on time. What I missed was the test loop: the first prototype showed the clamp marking the part, and every change meant waiting for the machine shop. I'd planned as if the first build would pass. I raised the slip in week five, not week eight, which helped production plan around it, but it was still a miss. Now I break a project into design, build, test and fix, and I assume at least one fix loop for anything new, with the shop's real lead time, not the one I hope for. My estimates are longer now, but they've been much closer. The trade-off is that I sometimes have to defend a longer number to managers.”

“At my last company, trailer chassis cross-members were cracking at the weld toe of a gusset after about a year. The design had been checked with a static load and a dynamic factor, and it passed comfortably. So I put strain gauges near the gusset on one trailer and logged a week of real routes. Rainflow counting showed far more medium stress cycles from rough roads than the design assumed, and when I added up the damage with Miner's rule against the S-N curve for that weld class, the predicted life was close to what we'd seen. The fix wasn't a thicker plate. I changed the gusset to a tapered end so the stress flowed in gradually, and moved the weld off the highest stress point. It added a little cutting time per part, but the redesigned frames have had no cracks in two years.”

“On a vibrating screen, the bolts holding the side plates kept loosening within weeks, and the site's answer was to add thread locker. I found the joint was slipping sideways under vibration. Once the plates slip, the nut can turn back on itself, and no locker holds that for long. Two things caused it. The bolts were short, so a tiny bit of embedding between painted surfaces lost a big share of the clamp load, and most of the tightening torque was going into friction, so the real preload varied a lot. I specified longer bolts with spacer sleeves so they stretched more, bare contact faces instead of paint, and torque plus angle tightening for a more consistent preload. We also marked the nuts to see any movement. Six months on, none had moved. The trade-off was a longer, more careful assembly step.”

“On a cast mounting bracket, my model predicted a peak stress well below the limit, but in the rig test strain gauges read almost twice what the model gave at the same point. I trusted the test, but I didn't just add a fudge factor. I checked my assumptions one by one. The model had the bolt holes fixed completely rigid, while the real bracket was bolted to a thin sheet panel that flexed. That changed the load path and pushed more load through one rib. I modelled the panel and the bolts with contact, and the strains came within a small margin of the gauges at all four points. With that correlated model, I then tested three rib layouts on screen and only built the best one. The trade-off was two extra days on the model, but it saved at least one more round of castings.”

“I start from the requirements and the design FMEA, and every test in the plan has to point back to one of them. Anything new or with a severe failure mode gets a physical test. Carry-over parts with a small change can go on correlated simulation. For sample size, I'm honest about what we prove. With a success run test, showing a reliability of 0.9 at a confidence of 0.9 takes 22 samples with no failures, and we usually can't afford that. So on my last project we tested fewer samples for longer, running them to a multiple of the target life, which can give the same demonstration with less hardware. We agreed that trade-off with the customer in writing. The catch is that it leans on an assumed Weibull slope for how the part wears out, so I take that slope from past test or field data, not a guess, and I check that any failed part looks like a field failure.”

“I wouldn't sign off, and I wouldn't quietly drop the failed sample either. First I'd tear down that sample and compare it with the others. Is there a build defect, like a missing weld or wrong heat treatment, that I can prove from the part? Or did it fail at the same spot the others were starting to show wear? Then I'd call the customer the same day, tell them what happened and when I'll know the cause. If it's clearly a build defect, I'd fix the build process, test replacement samples and show the evidence. If it's the design, the sign-off waits for a fix. If they need parts urgently, we might agree a limited release with extra inspection, but only in writing. The trade-off is a slipped date versus a failure in their hands.”

“We had five parts stacking up to set a clearance on a gearbox cover, each held to plus or minus 0.1 mm. Worst case, the gap could vary by plus or minus 0.5 mm, which was more than the seal could take. A root sum square stack gives the square root of five times 0.1 squared, about plus or minus 0.22 mm, which fit. But RSS only holds if every process is centred, roughly normal and independent. So I pulled capability data for each dimension from the supplier and our machine shop. Four were well centred with good capability. One, a cast face, drifted between tool changes. I kept that one on worst case, used RSS for the other four, and added a check of the cast face to the incoming inspection. We've had no seal leaks from that stack in over a year of production.”

“We had an aluminium gearbox housing with steel bearings pressed into it. It worked on the bench, but in a hot environment the outer rings started to spin in the housing. Aluminium expands about twice as much as steel. On a 100 mm bore with a 60 K rise, the difference in expansion is about 11 millionths per kelvin times 100 mm times 60, so roughly 0.066 mm. That was more than the whole interference we'd specified, so at temperature the fit went loose. We added steel sleeves cast into the housing so the bearing sits in steel, and tightened the fit a little. The trade-off was a heavier, more costly casting. Now every fit between different materials gets checked at the lowest and highest operating temperature, not just at room temperature.”

“After we raised the speed of an exhaust fan with a new drive, the bearing housings started vibrating hard. I took readings in three directions and looked at the spectrum. The big peak sat right at running speed, which usually means imbalance, but balancing the rotor only helped a little. So I did a bump test on the stopped fan and found the support frame had a natural frequency just above the new running speed. At the old speed we'd been well clear; now we were close to resonance, so even a small imbalance got amplified. I had two choices: slow the fan back down or stiffen the frame. We added two braces to the pedestal, which pushed the natural frequency well above the running speed, and the vibration dropped to a fraction of what it was. The trade-off was a short shutdown for welding.”

“The basic rating life assumes clean, well lubricated bearings with correct fits, and for a ball bearing life goes with the inverse cube of load, so doubling the load cuts life to about an eighth. In practice, most early failures I've seen weren't load at all. On a row of agitator motors that had just been moved onto variable speed drives, the bearings lasted a few months. I had the failed ones cut open. The races had fine, evenly spaced grey marks, like a washboard, which pointed to electric current passing through the bearing from the drive. The load zone was normal, so it wasn't overload or misalignment. We fitted insulated bearings at the non-drive end and a grounding brush on the shaft, and the next set ran well past a year. The trade-off was a more expensive bearing, but far fewer stops.”

“A new cooling water pump delivered well under its design flow and sounded like gravel. I measured suction and discharge pressure, flow and motor power, and plotted the point against the pump curve. The head was higher than the design calculation, so the real system curve was steeper than assumed; the installer had added extra bends and a fine strainer that was already partly blocked. The noise was the bigger clue. With the strainer losses, the pressure at the pump inlet was low enough that NPSH available dropped below what the pump needed, so it was cavitating. We cleaned the strainer, put in a coarser one with a pressure gauge across it, and removed two bends. Flow came back near design and the noise stopped. The trade-off was a strainer that needs checking every week.”

“I run the design FMEA while the layout is still rough, with manufacturing, service and quality in the room, because that's when a change costs nothing. On a hydraulic cylinder mount, one line was the hose being fitted to the wrong port during service. The effect was the machine moving the opposite way, so severity was the highest we rate, and our only detection was a label. Instead of adding a warning sticker, we changed the two ports to different thread sizes so the wrong hose physically can't fit. Occurrence went to almost nothing, and we re-rated the line in the next review. The FMEA also fed the test plan: lines with high severity got a specific test. The trade-off was one extra part number in the service kit.”

“At a plant I supported, gearbox couplings on the mixers were replaced every three months on a fixed schedule, yet they still failed between changes. I pulled two years of work orders and fitted the failure times. The pattern showed failures were spread out randomly, not bunched up with age, so replacing them on a schedule wasn't buying anything. Looking at the failed parts, most had worn unevenly on one side, which pointed to misalignment after motor changes. We stopped the fixed replacement, laser aligned every drive after any motor work, and added a quick vibration check on the monthly round. Coupling failures dropped to a couple a year, and we saved the labour of the scheduled swaps. The trade-off was training the fitters on the laser tool and making alignment sign-off mandatory.”

“Day one, I'd make sure the three customers are safe and have a workaround, and I'd ask service to get the leaking units or at least the seals and fittings back fast. I'd pull the build records for those serial numbers and look for what they share: a build week, a seal lot, a supplier change, an assembly station. If they share a lot, I'd put units from that lot in the warehouse on hold and check them. Meanwhile, I'd look at the returned parts myself, because a cut seal, a cracked housing and a loose fitting point to very different causes. By the end of the week I want a clear scope, a containment in place and a short update to management saying what we know and what we don't. I'd avoid promising a root cause before the parts are in my hands.”

“I don't start with the arithmetic. I ask the engineer to tell me the load cases and where each number came from, because most real errors are in what was left out. On a lifting frame, a junior engineer had a neat, correct calculation for the working load, but no case for the frame being lifted at an angle by one corner, which is how riggers actually handle it on site. That case put a bending load into a lug that was sized only for straight tension, and it came out under our standard. Instead of fixing it myself, I asked how the frame would be picked up in the yard, and he found it. We added the case and a thicker lug. He now writes a list of how each part is handled, transported and installed before he starts calculating.”

“A graduate joined my team who was strong on theory but froze whenever a design needed a judgement call, and brought every small question to me. At first I answered them, which made it worse. So I changed my approach. When she asked a question, I asked what she'd do and why, and I only stepped in if it was unsafe or expensive to get wrong. I also gave her one small assembly to own completely, including the supplier calls and the design review, with me in the room but not talking unless asked. The first review was rough, but she handled the next one alone. A year later she owned a full sub-system. The trade-off was that her first project took longer than if I'd done it, and I had to accept that.”

“I'd get both of them in a room and first agree what matters for this frame, before anyone argues for an answer: yearly volume, stiffness, weight, lead time to first parts, tooling cost and how likely the design is to change. Usually the disagreement is really about which of those matters most. Then I'd give them two days to fill in a simple comparison, including a rough quote for the pattern and the weldment. In my experience, a weldment often wins early, when volumes are low and the design is still moving, and a casting wins once the design is stable and volumes justify the tooling. If the data doesn't settle it, I make the call and write down the reason, so we can switch later if the volume forecast changes. The trade-off is that one of them won't get their way, so I'd tell them why directly.”

“Purchasing wanted to move a nylon gear from one moulder to a cheaper one using a different grade of glass filled nylon. The risks were strength at temperature, moisture uptake changing the size, and wear. I asked for material certificates and a first article inspection on every dimension, conditioned sample gears in humidity and remeasured them, and ran a gear rig test at the hottest operating temperature. I decided not to repeat the full product drop and noise tests, because those depended on the housing, which hadn't changed. The first samples swelled more than the old grade and tightened the backlash, so the moulder adjusted the tooth thickness in the tool. The second round passed. The trade-off was accepting some risk on long term wear, which we covered by pulling parts from early field units to inspect.”

“Our prototype pump housing was machined from solid aluminium, which was fine for ten units but took hours each. For a few thousand a year we moved to a die casting. That changed a lot. I added draft to every wall, evened out the wall thickness so it would fill and cool without shrink porosity, replaced thick solid sections with ribs, and added fillets everywhere the prototype had sharp inside corners. The biggest change was datums. On the prototype everything was machined, so any face could be a datum. On the casting, I set the datums on three small pads that get machined in the first operation, so every part sits in the fixture the same way, and after that only the bearing bore and the sealing face needed machining. The part cost dropped a long way per unit, but we paid for the tool up front and lost the freedom to change the design cheaply.”

“At my last company we moved about twenty thousand parts from shared folders into a PDM system. I led the mechanical side. The pilot on one product showed the real problem: the same bracket existed under three part numbers, and some assemblies pointed to old revisions that no longer matched what the shop was building. So before the main move, we ran a clean-up where each product owner confirmed the released revision against the latest shop drawings, and we merged duplicates with a cross-reference list so old numbers still found the right part. We froze engineering changes for one weekend and moved product by product. One assembly still lost its drawing links, and we fixed it within a day. If I did it again, I'd start the clean-up months earlier. The trade-off was a slow first month.”