I burn about 8 tons of a variety of hardwood pellets per season in my 2008 PB-105. I'm on my third burn pot and 3 ignitors (coincidental number of *3* failures, they were not coupled). My burn pot failures are first manifested by the classic bump formation at hole rows 4 thru 6 (from the open end), followed by a crack forming perpendicular to the hole rows as the season proceeds. The steel in the bump region is badly spalled probably due to heat of the burning gases which are given off during chemical decomposition of the wood. Interestingly wood does not burn, it gives off gas which if heated sufficiently will ignite. The gas released from the heated pellets is comprised mostly of hydrogen sourced from cellulose (C6H10O5) and lignins (C9H10O2, C10H12O3, C11H14O4) in the wood, mostly in the form of sinapyl alcohol and coniferyl alcohol. If you carefully observe the ignition process through the window in the fire box door of the PB-105, you will eventually see smoke pouring out of the burn pot (I have the finned ignitor), it helps to use a flashlight to observe the smoke. The smoke is the visual indicator that gasses are being liberated from the pellet fuel. At a sufficiently hot temperature ignition happens all at once (sometimes you can hear an audible *pop*) and the burn pot lights and flames start shooting upwards towards the internal boiler cross pipes. The pellets in back of the flame front from the auger output to the first row of holes are unburned. If you empty a lit burn pot you can observe these pellets that are effectively cued up to be exposed to the heat and air required to release their combustion gases. Also, if you have a entire burn pot full of pellets during the ignition process you can appreciate that what will happen is that a great deal more gas will be liberated due to the increased quantity of pellets and when ignition occurs, there is significant overpressure created by the sudden increase in temperature from the recently combusted expanding gas P=nRT/V which causes the spring loaded backfire plate to compress the retention springs to their minimum height, allowing the overpressure gas to safely be discharged along with an unhealthy gulp of smoke, into the boiler room. Backfires can be prevented by setting the dip switches 1,2, and 3 to minimize the pellets in the burn pot during ignition, as mentioned in several previous threads. Minimizing the pellets during ignition is key to long term reliable operation especially as the outside temperature begins warming up and the cycle times extend, often using up all the fuel in the burn pot which required a re-ignition cycle to commence as often as several times per day.
Steel is definitely not fire proof, it degrades significantly in the presence of high heat. I found a graph of structural steel (I couldn't find one for boiler steel which probably has a higher temperature performance) strength vs temperature which clearly shows how rapid steel strength drops off as a function of temperature, note temperature is in °C. By 600 °C steel only has 1/2 its strength and loses most of its strength altogether at about 1200 °C.
In the presence of high air flow (air has 21% oxygen which reacts with CO liberated during combustion of HxCyOz molecules creating CO2), wood pellets can achieve burn temperatures of 1000 to 1200 °C and it shouldn't be a surprise why the steel in the burn pot degrades. Steel has a thermal coefficient of expansion of about 13 ppm/°C. If the outer edges of the burn pot are at a much colder temperature, say 300 °C, the difference of 700 °C x 13 ppm/°C causes an expansion of approximately 0.05 inches across the area of the burn pot where the burning gases are most intense and since the temperature of the steel is at or above the plasticity limit for the steel, the steel expands and as we know if has a propensity to expand upwards into the flame because that side of the steel is warmer than the steel in the plenum side. As the steel rises into the burning gases, it gets even hotter and since the steel has exceeded the plasticity temperature it expands and permanently deforms until finally the expansion has reached equilibrium. At this point the steel is subjected to repeated flame fronts that beat against the surface of the steel like a blow torch (remember those things?) causing the surface to spall and exhibit macroscopic degradation until finally a crack is formed and I suspect in the limit a hole will form as the material at the bump site is completely compromised and falls away. Normal high temperature steel will likely always exhibit this failure mode so it will take a very high temperature specialty steel to accommodate the high temperatures that exist in the burn pot. I would like to take in situ measurements of side walls and the bump burn pot temperature to confirm these calculations and will write back with that information as well as anything I can find on very high temperature steel. It is plausible that the good folks at Harmon don't yet fully understand the root cause of this failure and therefore are shooting somewhat in the dark for a solution which is why Harmon has not simply *fixed* the burn pot problem.
Upon further reflection I now believe the temperature gradient from the inside of the burn pot floor to the inside surface facing the plenum is worse than originally thought. The combustion air is near 20 °C and is percolating up through the holes in the burn pot floor creating a temperature gradient closer to 975 °C, thereby making the thermal expansion worse than the 700 °C temperature difference I used in earlier calculations to determine the expansion of the burn pot floor. The high thermal gradient is undesirable and can be mitigated by pre heating the combustion air. Just thinking about it overnight, it may be possible to construct the burn pot with a double floor between the fire bearing surface and the plenum. The space between the two floors creates a volume where the combustion air can pick up some heat before entering the burn pot surface. Clever routing/baffling of the incoming air between the double burn pot floor, may allow it enough time to pick up adequate heat to lessen the thermal gradient. I don't know for sure if this would work, but by constructing a SolidWorks model and with a few hours to tinker around, it should be possible to figure out exactly what is happening and then try different configurations to lessen the thermal gradient effects. I measured 923 °C at the surface of the burn pot using a K type thermocouple probe. I posted this to the Ectoteck thread before I realized it wasn't a PB 105 thread.