Every other year, a well known large Materials Science & Engineering conference (TMS) takes place and students compete to construct the best sword.

In our department of Materials Science & Engineering at McMaster university, our steel research is one of the best in the world because of Hamilton's historical ties to the steel industry. We had tons of furnaces, but not many expendable ones where a bunch of undergraduate and graduate students could play around.

Cue the miraculous idea to create our own furnace!

The build is summarized in this video for our entry into the TMS Bladesmithing competition. I would recommend you watch the video instead of my copy-pasted furnace blurb from our report below.

In case you are wondering, we did not win :(

Design of Furnace

An open-ended coal and coke powered box furnace was constructed to heat the sword to an austenitizing temperature for forging. One end of the furnace was used for live coke and coal refueling whereas the other end was used to bring the sword in and out of the furnace. The dimensions of the outside of the furnace was 0.46 x 1.68 x 0.93 m whereas the inner dimensions were 0.17 x 0.24 x 0.93 m. The dimensions were chosen based on the dimensions of the bricks used for construction.

Air was supplied to the furnace using perforated steel piping which was connected to a 100 PSI airline. Only < 5 PSI of air was required to reach temperatures above 1473 K. The highest temperature recorded in the furnace exceeded 1800 K when a steel sample melted.

The furnace walls were constructed as a bilayer which creates an effective composite structure with superior structural and insulating properties. A structural brick (k = 0.385 W/m-K) and an insulating firebrick (k = 0.17 W/m-K) was used to line the furnace walls where the structural brick was the load-bearing wall which supports the roof and chimney. Material properties (bulk) were taken from Cambridge Engineering Selector 2016. The bricks on the ceiling of the furnace were supported by a steel grate underneath to prevent the bricks from falling. A sheet steel chimney was used to bring the off-gases to the buildings internal fume-hood ventilation system.

Full view of the finished furnace with the chimney connecting to the fume-hood.

To find an optimal furnace size to heat the billet for forging, a two-fold approach was used. First a preliminary size of the box furnace was constructed in AutoDesk Inventor 2016 then the CAD file was imported into the Finite Element Analysis (FEA) software ANSYS 16.2. FEA was used to troubleshoot where weak aspects of the design might arise from.

The following parameters were used to construct the simulation:

Temperature of air outside of furnace – 298 K
Surface and film temperature of outer furnace walls – 402 K and 350 K
Natural convection on the vertical outer walls (vertical plate) – 6.32 W/K-m2
Natural convection on the outside roof (horizontal plate) – 0.9 W/K-m2
Velocity of airflow inside the furnace – 15 m/s
Forced convection on the inner walls – 12 W/K-m2
Fixed surface temperature of the inner walls – 1173 K
Firebrick – k = 0.17 W/m-K, cp = 750 J/kg-K, ρ = 550 kg/m3
Insulating Concrete – k = 0.385 W/m-K, cp = 980 J/kg-K, ρ = 1150 kg/m3
Structural Steel – k = 60.5 W/m-K, cp = 434 J/kg-K, ρ = 7850 kg/m3

Radiation losses of the ceramic walls was found to be negligible and no convective losses were assumed on the bottom face of the furnace which was touching a table.

Final FEA model with the same dimensions and properties as the physical furnace constructed. a) Surface temperature isotherms b) Total heat flux at each element

To determine the natural convection coefficients, the properties of air (kinematic viscosity, thermal conductivity, the Prandtl number, and the thermal diffusivity) were evaluated at the film temperature (assumed). These properties combined with properties from the furnace were used to evaluate the Ra number, followed by the Nu number then finally solving for the convective heat transfer coefficients. The specific empirical Ra and Nu number formulas were found according to the geometry required (flat plate and vertical plate). To determine the inner forced convection coefficient, the Re number was found assuming the inner airflow velocity was 15 m/s. Using this value with the Nu number yields the forced convective heat transfer coefficient on the inside of the furnace and the chimney.

With these constants found, the simulation was initialized in the Transient Thermal package within ANSYS 16.2. To speed up thermal equilibration, the initial temperature was assumed to be 473 K. After imposing a constant surface temperature of the inner furnace walls to be fixed at 1173 K, the simulation converged after 50 000 time-steps. The long equilibration time was required due to instabilities near the start of the simulation.

The global minimum of the simulation shows convergence after ~50,000 time-steps. The location of the global minimum was found on the outside corner of the box furnace.

To improve areas of the furnace, the flux and the temperature was visualized using vectors and contouring isotherms respectively (figure above). It was found that there was significant heat loss through the roof of the furnace so the roof insulation was doubled. In addition to this, bricks were placed at the entrance and the exit of the furnace to reduce heat losses. The steel chimney was found to be an incredible heat sink for removing heat from the furnace, so insulating brick was stacked as high as possible around the chimney to limit this. The “hot-zone” where the coke and coal was the hottest was moved between the furnace entrance and the chimney to avoid heat losses from the chimney.

The student version of ANSYS severely limits the total number of elements allowed to be utilized. Using fewer elements results in less accurate simulations and longer equilibration times. A total of 2806 elements and 17 616 nodes were used for the simulation. The 1.3 m chimney was omitted from the FEA calculations due to the amount of elements required for meshing it.