Heat Transfer Design Challenges

Overview

Two open-ended design challenges from EML4140, the heat transfer course. Each one was set by a company operating near Gainesville, and each came with a real thermal duty, a real budget, and real constraints instead of numbers picked to make the math come out clean.

That changes how you work. The answer is not a single number at the bottom of a page. It is a configuration you have to argue for, sized from the governing correlations, modeled as something that could actually be built, and costed to run for a year. Both were done in a five-person team.

GE Vernova, steam condenser for a 110 MW power plant

GE Vernova builds roughly a quarter of the world's electric power generation equipment. The brief was a conceptual steam condenser for the John R. Kelly Generating Station, the combined cycle plant that Gainesville Regional Utilities runs in town.

The condenser is one of the largest heat exchangers in a power plant and it sets the steam turbine's backpressure, so it drives the efficiency of the whole Rankine cycle. Get it wrong and the plant burns more gas for the same megawatts.

What we were given

350,000 lbm/hr of saturated steam leaving the turbine at 120 °F, condensate out at the same temperature, ambient air at 80 °F and 74% relative humidity, a 7 °F cooling tower approach, and 8,000 operating hours a year.

What we chose

A horizontal two-pass shell-and-tube surface condenser. One shell, steam condensing outside straight tubes, cooling water inside. The decision that shaped everything else was reusing the plant's existing four-cell Marley 600 crossflow tower, which avoids buying a new tower and avoids the footprint and the fan power an air-cooled condenser would have needed.

Tubes are ASTM A249 TP316L, 1.000 in outside diameter on a 0.028 in wall, laid out on a square pitch at 1.25 diameters so there is a clear lane for mechanical cleaning. Target tube velocity is 7.5 ft/s, fast enough to keep silt and fouling from settling, slow enough to avoid erosion-corrosion of the 316L.

How it was sized

Psychrometrics first, because the wet bulb is the lowest temperature the tower can ever approach and it anchors the entire water loop. That worked out to 73.54 °F, so cooling water enters at 80.54 °F and leaves at 95.54 °F across a 15 °F range.

The duty is pure latent heat, since the steam enters as saturated vapor and leaves as saturated liquid at the same temperature. Tube count is set by velocity rather than by heat transfer, which gives 5,882 tubes in two passes of 2,941. A Reynolds number of 69,806 confirmed the water side was fully turbulent before applying Gnielinski for the film coefficient, and the series resistance network produced an overall coefficient of 1,755 W/m²K.

Where it landed

• 105.18 MW thermal rejected
• 48,108 gpm of cooling water
• 37,007 ft² of heat transfer surface, 24 ft tubes in a 28 by 10.6 ft footprint
• 1.186 MW of auxiliary pump and fan power
• 11.75 kPa absolute condensing pressure, which keeps turbine backpressure low
• Major material capital cost between $0.43M and $0.77M
• About $1.19M a year to operate on GRU's published non-residential rates

The tradeoffs are real and the deck says so. Reusing the tower ties the design to water availability and to permitting limits. Fouling degrades performance and has to be cleaned on a schedule. It needs a vacuum system to pull non-condensables, and it needs physical access to pull the bundle for cleaning and retubing.

Modeled by Evan Alvarado. I worked with him on optimizations and on workarounds for the constraint analysis.

Team presentation, GE Vernova condenser

Sandvik, oil cooler for a rotary drill rig

Sandvik Mining and Rock Technology builds large surface mining drill rigs in Alachua, just outside Gainesville. Their xSeries and iSeries rotary drills throw off about 25,000 BTU/min, or 439.6 kW, of waste heat from the hydraulic and lubrication systems, which a fan-box oil cooler currently handles. The brief was to find a meaningfully different way to take some or all of that load.

What makes it interesting is that the oil has to stay inside a window rather than simply get as cold as possible. Too cold and viscosity climbs until it is hard to pump. Too hot and the oil degrades, vapor forms, and the pump cavitates. The ceiling is 265 °F, the target is 220 °F, and the machine has to work in ambient air anywhere from −40 °C to +50 °C.

The concept

A two-fluid concentric cooler. Hot oil runs down an inner pipe and gives up heat across the wall to a 50/50 ethylene glycol and water coolant in the surrounding annulus, running counter-flow so the temperature difference stays as large as possible along the whole length. The outer pipe is finned and sits in front of the rig's existing suction fan, so the air the fan already moves carries the heat away. Four parallel runs, each with two 180 degree bends, twelve passes in total.

How it was analysed

As a thermal resistance network, assuming 1D radial conduction and convection, steady state, no internal generation, and turbulent flow in the pipes. Breaking it into resistances in series meant each link could be sized on its own and then summed.

The fin array carries most of the geometry. Single-fin efficiency came out at 0.749 and overall surface efficiency at 0.762, across 4,456 fins giving 51.6 m² of finned area. That produced a conductance of 823 W/K for one 3 m unit, and 3,292 W/K for the four running in parallel.

Where it landed

• 12,530 BTU/min rejected in the worst case, with air at +50 °C
• 29,390 BTU/min in the best case, with air at −40 °C
• Oil temperature drop of 30.9 °F to 63.8 °F across the configuration

The best case is also a failure mode. At −40 °C the design pulls more heat out than it should and risks overcooling the oil, which is the viscosity problem from the other direction. The recommendation is a variable speed fan reading ambient temperature sensors and backing off when it is cold, rather than a fan that runs flat out regardless.

The resistance breakdown is the most useful result in either project. The oil-side film is 35.8% of the total and the fin and air side is 34.9%, so together they are about 70% of the resistance. Conduction through both pipe walls is 0.4%. That says plainly that a thicker or more conductive pipe buys nothing measurable, and every improvement worth chasing is either in the oil film or on the air side.

One honest limitation. The three convection coefficients were assumed from reference tables and empirical data rather than derived, because deriving them needs material from later chapters than the course had covered at that point. The deck states the assumption openly and lays out the method that would be used to calculate each one properly, which is the right way to handle a gap you know about.

Modeled by Evan Alvarado. I worked with him on optimizations and on workarounds for the constraint analysis.

Team presentation, Sandvik oil cooler

What I took away

Both challenges made the same point from opposite directions. The heat transfer math is the easy half. The hard half is deciding what to optimise, and that only becomes clear once you know where the resistance actually sits or what the thing costs to run for a year.

On the condenser, reusing an existing cooling tower mattered more to the economics than any refinement to the tube layout. On the oil cooler, the resistance network said outright that the pipe wall was irrelevant and the air side was half the problem. Neither of those falls out of a textbook problem, because a textbook problem already tells you what to solve for.