A factory whose main products are paints and varnishes was using a single stage zeolite filter to treat their VOC-contaminated waste air streams. The filter was thermally regenerated over night. Sometimes, the zeolite bed was found to be warm in the morning, which caused no concern since it was assumed that the filter simply had not been cooled enough after regeneration. One morning, however, the temperature did not go down after the air stream was turned on again, but instead it went up – well above 800 °C, leading to discoloration of the steel and a total loss of the zeolite filling.
Why thermal runaway occurs
Generally speaking, a thermal runaway will happen if (a) heat production in a unit is larger than heat loss and (b) heat production rises with the temperature of the process.
The latter is mostly the case with chemical reactions, which follow Arrhenius‘ Law: Reaction turnover, and therefore heat production, rises exponentially with temperature. On the other side, in insulated adsorption filters the only significant contribution to heat loss is via the air stream. In this case, the air stream was being turned off over night, leaving the filter with no cooling at all.
What had happened?
The filter was equipped with a combined thermal desorption and condensation unit, powered by a chiller. The chiller was not always working optimally, which led to sometimes very incomplete desorption of the zeolite. This, too, was known, because it lead to early breakthrough of the filter on the following day.
The partially cooled, partially laden, closed-off zeolite filter had several hours to develop hot-spots. A diffused and slow oxidation of the adsorbate is to be expected on many zeolites, some being more catalytically active than others. Mostly these reactions are very slow, lead to no noticeable temperature rise and are extinguished once a cooling air stream is blown through the zeolite bed. Occasionally, though, the chemical reaction in one spot will be slightly more active, leading to localized temperature increase. This heat does not diffuse, because there is no convection in the filter and zeolite is thermally very insulating. At the warm spot, the chemical reaction speeds up, leading to higher temperatures, which in turn speed up the chemical reaction. A hot spot has formed, limited only by the speed of oxygen diffusion to the site. Once fresh oxygen is let in the next day, the cooling effect might not be sufficient to quench the oxidation, but the opposite happens: The oxidation reaction flares up, takes over the whole zeolite filling and stops only after the combustible organics are completely burnt.
What to do?
The operators wanted to repair the filter and use it again, as it generally fulfilled their requirements. Of course, they wanted to make sure that such an event would not happen again. The insurance company also demanded a convincing strategy for future fire prevention. We did the following steps:
- Chemical analysis of the burnt adsorbent could do nothing in this case to elucidate the course of events. I did some theoretical calculations to make sure the numbers matched, that the thermal runaway could indeed have been the result of autocatalytic oxidation.
- While all hydrophobic zeolites – the kind you need for adsorptive waste gas purification – have a certain amount of catalytic activity, they are not equal in this regard. We did differential scanning calorimetry on several laden samples of different zeolites to find the type with the lowest catalytic activity.
- We added some monitoring to the filter – in this case, carbon monoxide sensors and two more temperature sensors were built in. Automatic responses to the detection of hot-spots, such as turning up the blowers or flushing the filter with nitrogen, were programmed into the process control software. It took several days of measuring under real operating conditions to determine the normal baseline fluctuations, so as to not make the automatic quench overly sensitive.
- We determined a new shutdown regimen – correct responses to incomplete regeneration, a constant minimal air flow through the filter and a nitrogen flush before total shutdown were added to the process control software.
The insurance company accepted the retrofit and the fire prevention concept, and under the new regimen, the filter has never again shown signs of any undue temperature increase.
I am taking a closer look at the burnt zeolite in this blog article (in German): „Wenn Zeolithe brennen„.