Ceramic Industry

Kiln Connection: Always Question the Data

November 1, 2002
The late G. Bickley Remmey, founder of Bickley Furnaces, Inc., was a great innovator in the kiln and burner design field. I look back on his way of thinking often when I am working on a technical problem, and I try to imagine how he would solve the problem.

The late G. Bickley Remmey, founder of Bickley Furnaces, Inc., was a great innovator in the kiln and burner design field. I had the opportunity to work with him on a project in Ohio in 1978, and his clarity of thinking and ability to see and solve problems was remarkable. I look back on his way of thinking often when I am working on a technical problem, and I try to imagine how he would solve the problem. This process sometimes helps me to construct solutions, and I hope it will help you also.

The Firing Problem

For that project, I was working on a new kiln to fire lightning arrestors, which are large electrical insulators that are about 18 in. in diameter and 16 to 18 ft high. These shapes were set free-standing on a large kiln car and then fired in a shuttle kiln. The problem that we were working on seemed to be a tough one—while the kiln had excellent temperature uniformity, the insulators often fell down during firing. We didn’t know exactly at what point in the cycle the pieces fell, because the kiln temperature uniformity was so good that you could not really see any of the parts in the kiln at red heat.

The firing cycle was based on a laboratory analysis of the dilatometer curve. Developed from a “known curve” that was in use in an older kiln, the cycle was very conservative, with slow heating rates through the quartz inversion and the oxidation and shrinkage periods.

Wild Goose Chase

We started to check everything. We looked at the kiln car and found that the setting surface warped somewhat during heating due to flue steel components, so we field modified the kiln car drastically. The car stopped warping, but the pieces kept falling.

Next we focused on the combustion system. On a rush basis, the burners were set up in our laboratory, with piping identical to the kiln in Ohio, and tested. Modifications to the gas nozzles were made to improve mixing and shorten the flame length. At the same time, the ratio control system was altered. These “fixes” were made on an emergency basis in the field. The burners looked better, but the pieces kept falling.

Someone then suggested using more excess air, with the thought that higher levels of air would promote better temperature uniformity. Thus, we changed the secondary air program, once, twice and then a third time. We managed to nearly double the fuel consumption (!), but the pieces kept falling.

To continue to improve the temperature uniformity, we placed monitoring thermocouples in front of each burner as precisely as possible and recorded the temperatures as the kiln approached the shrinkage period of the ware. During this effort, we made very slight adjustments in the fuel input to each burner in each zone to achieve identical burner jet temperatures at each burner location. Because of the burner and ratio modifications made earlier, we achieved excellent conformity of the temperatures of gases emanating from each burner. But unfortunately, the pieces still kept falling.

The Breakthrough

The customer was starting to lose patience, and our company had spent tens of thousands of dollars in field expenses. My boss began to think in terms of the shrinkage rate of the ware and conceived a test where the entire kiln was converted into a dilatometer. Long mullite tubes were obtained, and they rested on two insulators. Because the kiln was so tall, we rigged up a cable system so that we could monitor the change in length of the ware. Though our apparatus did have some limitations, the lineal shrinkage of the pieces was so large (15 to 16 in. total) that we had a very accurate system. At the same time, a large piece of ware was “wired” with a dozen thermocouples, from top to bottom, to establish the true temperature of the product. We ran the test, at least until the ware fell over, and then Mr. Remmey came out to Ohio to visit.

The Analysis

The thermocouples that we installed indicated excellent temperature uniformity in general. We found that the delta T in a single piece had little variation over its length, and that the delta T across the piece diameter ranged from 5 to 60 degrees F, which was totally dependent on the kiln heating rate.

The full size dilatometer data was revealing; we found that the shrinkage of the piece started about 75 degrees F earlier than the lab dilatometer curve. We checked—and rechecked—our data. Mr. Remmey looked at all of the information and made the following observations:

  • The laboratory dilatometer curve was run at a rate of 240 degrees F/hour, which was much faster than our kiln heating rate at the initial shrinkage point. Accordingly, because of our slower kiln heating rate, the shrinkage of the production pieces occurred earlier than expected.
  • Our kiln program slowed down too late to achieve the best uniformity during shrinkage, since we were mislead by the laboratory data. And the older kiln that served as a model for the firing cycle likely had a larger difference between indicated kiln temperature (hotter) and ware temperature (cooler), which further misled us.


On the 28th cycle of the kiln, we simply reduced the rate of heating at an earlier point in the cycle, based on our real shrinkage data. And from that cycle on, there were no more problems. While all of our work on the cars, combustion system, etc., probably helped the problem, none of these was the solution. One simple adjustment was all it took to fix the problem.

Looking back, the solution seems fairly obvious. The biggest barrier to finding it turned out to be the most scientific information—the lab dilatometer curve. Even though it was accurate, it did not apply exactly to the product being fired in production because of heating rate differences.


Today, in my troubleshooting efforts, I always ask for information—and then I question its source. More than half of the time, the initial data provided to me is incorrect or does not apply. Conclusion: Always question the data, no matter who provides it. Bad data is worse than none at all, since it will always lead in the wrong direction.