I don't think this is possible.
All TI's use what is called a bimetallic strip. Essentially two metals bound together and a strain gauge is used to measure the amount of distortion causes due to the differences in thermal expansion. That being said, there is a limit to how much temperature difference you can tolerate as while the expansion is natural and causes no strain, the resistance due to the other metal trying to pull or push the first one causes some large stresses; at some point your strip just fails as either the connections between the metals or the metals themselves break. Based on what is available, I'd estimate that range of feasibility is maybe half the temperature range that you are suggesting.
I would suggest utilizing the expansion of the furnace itself in order to provide some sort of marker on the floor or something if able; I assume the furnace expands quite a bit from 0 degrees to 1200 degrees, in the order of inches. When it is heated from 0 C to 1200 C, the typical piece of steel would expand somewhere between 0.01 to 0.0144 meters/meter of length. or 10 to 14 mm per meter of length, assuming a coefficient of thermal expansion of around 10 to 12x10^(-6) m/(m K). If your appliance is a couple meters long you might have up to 3 inches of possible expansion to work with.
I also imagine you might be able to find hydraulic temperature valves to operate some sort of pneumatic system (IE, a globe valve of sorts with an expanding stem.) although that is just me guessing.
If I might ask, what is the rational behind not using energy in the building? Certainly there is tons of energy being used considering that the process is being heated some 1200 degrees! I'd find it hard not to justify the energy required for proper operations alarms, personally.
Edited by MTumack, 18 April 2017 - 11:08 AM.