Presented at the Munitions Technology Symposium IV Conference, Reno, NV, February 1997.

Lessons Learned Developing a Pyrotechnic/Thermite Torch

Alan C. Munger
Barry T. Neyer
EG&G Optoelectronics / Star City
Miamisburg, OH 45343-0529

Contact Address
Barry T. Neyer
PerkinElmer Optoelectronics
1100 Vanguard Blvd
Miamisburg, OH 45342
(937) 865-5586
(937) 865-5170 (Fax)


EG&G Optoelectronics manufactures a pyrotechnically driven Thermite Torch that can be used for a variety of purposes. For example, it will effectively destroy a small electronic package by incineration. The device can be reliably ignited by a standard 3.5 amp 2-3 ms constant current pulse or a CDU charged to 20 mJ.

The device has been subjected to various environmental conditions, and destructively tested cold, ambient, and hot. It always met the minimum requirements for burning a hole in a stainless steel plate, and had minimal variation in the hole diameter.

Although the final design proved to be very satisfactory, the initial design had various design flaws that were not discovered until Qualification. The design was developed by the customer. EG&G Optoelectronics was required to meet an extremely tight production schedule, and was not contracted to thoroughly test the design. This paper will discuss the Thermite Torch design, its performance, and the problems that were encountered during Qualification. It also discusses lessons learned from the design and qualification approach.

Technical Papers of Dr. Barry T. Neyer


EG&G Optoelectronics manufactures a Thermite Torch that can be used to supply a compact heat pulse to an object. These types of devices could be used for a variety of purposes. For example, these devices could be used to destroy sensitive components to prevent their capture by adversaries. They could also be used as a source to ignite a variety of materials.

Figure 1: Electric Match

The device consists of a small Electric Match containing eleven milligrams of Ti:KClO4 pyrotechnic. Four milligrams of fine particle pyrotechnic blended in the ratio of 33:67 weight percent was loaded against the bridgewire. Seven milligrams of coarse particle pyrotechnic blended in the ratio of 41:59 weight percent was pressed in a second load. Figure 1 shows a schematic diagram of the match. The Electric Match portion of the device is a stand alone device that could serve a variety of ignition functions. Thus, this device was tested and qualified independently of the Thermite Torch. The number of tests required by the qualification program for the Thermite Torch itself is reduced because many of these requirements are satisfied by meeting the qualification requirements of the Electric Match.

Figure 2: Thermite Torch (Original Design)

The hot burning particles of the pyrotechnic ignite a high density CuO:Al thermite pellet that is hot pressed to a density of approximately 4.0 g/cc. The blend ratio of the thermite is 82:18 weight percent. This ratio is chosen so that all of the oxygen will recombine with the aluminum, resulting in maximum energy output from the device. A schematic diagram of such a device is shown in Figure 2.

The theory of operation of the torch is quite simple. A constant current of several amps will heat the bridgewire in the Electric Match, and ignite the Ti:KClO4 pyrotechnic. Alternatively, the match can be ignited by the discharge from a Capacitance Discharge Unit (CDU) charged to 20 mJ. The pyrotechnic quickly burns and ruptures the closure disk of the Electric Match. The hot gas and particle stream rushing out of the Electric Match ruptures the thin region of the thermite pellet. The ignited thermite passes through the output hole, rupturing the torch output disk. The molten copper expelled from the device is then able to burn through the object placed near the output port. (In the case of steels, the molten copper readily penetrates the steel grain boundaries and causes failure and penetration.)

Qualification Tests

The Thermite Torch was subjected to a rigorous qualification program. Both the Electric Match and the Thermite Torch experienced Thermal Shock (-54 C to 74 C), Mechanical Shock (200 g, all axes, 3 ms, half-sine), and Vibration (10 g, 10-2000 Hz, 60 minutes) environments.

The Electric Match passed 1/8 Amp No-Fire Tests, Output Pressure Tests, Resistance After Fire Tests, and a number of No-Fire and All-Fire Tests. Some of these devices were tested after experiencing the environments mentioned above, and some were taken from a control group. The Thermite Torches were tested at cold, ambient, and hot temperature in both the constant current and CDU firing mode.

Threshold Tests of the Electric Match

The Electric Match passed a 0.9 Amp 5 minute constant current No-Fire Test and a 3.5 Amp 15 ms constant current All-Fire Test. The match also passed three 40 mJ All-Fire CDU tests; one group was tested cold after experiencing the environments mentioned in the previous section, two groups were tested at ambient temperatures, one after experiencing the environments and a control group. All tests were conducted using the Neyer D-Optimal Test [Neyer 1994] procedure using a sample size of 20 units. Table 1 summarizes the results. The Mean and Standard Deviation are the maximum likelihood estimates derived from the statistics. The No-Fire and All-Fire Levels listed in the table are computed at 50% confidence. The table shows that the Electric Match meets the threshold requirements.

Figure 3: Constant Current All-Fire Test


Table 1: Electric Match All-Fire and No-Fire Test Summary

Test Requirement Mean Standard Deviation All(No)-Fire Level
Constant Current No-Fire >0.9 Amps @ 5 min 1.31 Amps 0.089 Amps 0.95 Amps
Constant Current All-Fire <3.5 Amps @ 15 ms 1.96 Amps 0.064 Amps 2.22 Amps
CDU All-Fire After Environments (Cold) < 40 mJ 13.5 mJ 0.81 mJ 16.8 mJ
CDU All-Fire After Environments < 40 mJ 10.4 mJ 0.51 mJ 12.6 mJ
CDU All-Fire Without Environments < 40 mJ 13.9 mJ 1.05 mJ 18.2 mJ

Figures 3 and 4 show the results of the threshold test for both constant current and CDU firing modes.

Figure 4: Capacitance Discharge Unit All-Fire Test (Cold after Environments)

Destructive Testing the Torch

The Thermite Torch Test Program required testing at three different temperatures (-54 C, room temperature, and +74 C) with both the constant current and the CDU firing systems. Eight units were tested with all six combinations of temperature and firing mode. Half of these units experienced the thermal shock, mechanical shock, and vibration environments mentioned above, and half experienced no environments.

The output of the torch was required to produce a hole greater than 0.100 inch in diameter in a 0.0625 inch thick stainless steel plate when the stand-off from the plate is 0.040 inches. The output from the devices tested is shown in Table 2. The table shows that the diameters of the holes created in the steel plates varied extensively. Several of the holes were just slightly over the required diameter, one was under, and two devices (the tenth and nineteenth tested) failed to produce any holes at all. The smaller holes were also irregular in shape. The two torches that did not produce any holes were discolored and quite warm to the touch.

Dissection of the first device to fail showed that the thermite was partially burned, but that the output did not break through the output disk. Tests continued to determine if more devices would fail the output requirements. The fourteenth unit had low output, and the nineteenth unit also failed to create any whole. After the third failure, we felt confident that we had a systematic failure, and would be able to determine the root cause.

Table 2: Hole Diameters Produced by Thermite Torch (Original Design)

Firing Temperature Energy (mJ) Hole Diameter (in) Firing Temperature Energy (mJ) Hole Diameter (in)
Ambient 19.8 0.137 Ambient 21.8 0.148
Ambient 19.3 0.158 Ambient 22.2 0.101
Ambient 19.6 0.110 Ambient 20.2 0.109
Ambient 20.5 0.108 -54 C 88.6 0.143
-54 C 98.8 0.143 -54 C 82.2 0.000
-54 C 94.5 0.144 -54 C 78.0 0.140
-54 C 83.8 0.127 -54 C 62.0 0.090
-54 C 89.2 0.133 -54 C 88.8 0.139
-54 C 76.7 0.000      

Failure Mode Found

When the failures were discovered, EG&G Star City investigated all of our processes to see if there was any way that we could have violated the device build procedures. An exhaustive search showed that all of our processes were well within tolerance limits, and that process variation much larger than allowed would nevertheless produce reliable devices.

The only process that did not have substantial margin in the design was in the thickness of the thermite material in the thin region. Review of the data from the small pilot production lot showed that the web thickness was kept at less than 10 mils, while the requirement was relaxed to 20 mils for production. Thus, sufficient testing was not performed to ensure that the increase in the web thickness would not cause a problem.

Inspection of the dissected unit showed that an appreciable amount of thermite material filled the exit hole of the Thermite Torch. Thus, we concluded that some of the torches failed either because the output from the Electric Match was insufficient to break through the web, or because the output did break through, but that the extra thermite material in front of the flying plate plugged the exit hole. The web was originally designed into the pellet because it was thought that breaking a thin portion of the thermite would expose sharp edges, resulting in easier ignition. It was also thought to be desirable to avoid any pinch points during the thermite pressing operation.

The New Design

To test whether too thick of a thermite web was the source of the problem, we quickly modified several of the torches by removing the closure disk, drilling an 0.080 inch hole through the thermite, and welding a new closure disk on the Thermite Torch. Figure 5 shows the modified design. Because this design did not have any thin edges that the original designer thought were needed for ignition, a substantial concern was that the Electric Match might not reliably light the new design. Thus, we also built several new Thermite Torches with a 0.150 inch hole through the thermite.

To ensure that other aspects of the design were not also on a parameter edge, we also built devices with an extra output disk, requiring the torch to cut through an output disk twice as thick. During these "re-development" tests, we also varied the test conditions to determine if different Electric Match firing conditions had any effect on the Thermite Torch output.

Figure 5: Thermite Torch (Final Design)

"Re-Development" Tests

A total of sixteen modified Thermite Torches were tested. Half had a 0.080 inch hole, and half with the 0.150 inch hole. Three of the torches had two closure disks; half of the devices were environmentally conditioned prior to testing. Half of the devices were tested cold, the remaining at ambient. Half of the devices were tested in the constant current mode, and half were tested in the CDU firing mode. The result of all the tests was that all devices functioned to specifications. The average hole diameter in the stainless steel plate was 0.138 inches, with a standard deviation of 0.010 inches. Thus, we concluded that the "fix" not only cured the torch failure problem, but that it also made the performance of the device much more consistent.

Thermite Torch Final Design Qualification

The Thermite Torches were all reworked the same way as the "re-development" units; by removing the closure disk, drilling an 0.080 inch hole through the thermite, and welding a new closure disk on the Thermite Torch.

Because there was no question about the qualification of the Electric Match, we were only required to repeat the Qualification Tests for the Thermite Torch. Ten Thermite Torches were tested to each of the same six combinations of firing modes and temperatures mentioned above. Four of the torches in each group were first subjected to the environmental conditioning. The results are summarized in Table 3. The table shows that all combinations of temperatures and firing modes produced consistent outputs. The standard deviation for the reworked devices was much smaller than for the previous devices, indicating that the new design was far superior to the previous design.

Table 3: Torch Qualification Test Summary

Test Average Diameter Standard Deviation
3.5 Amp Constant Current @ 15 ms, Cold 0.136 0.0027
3.5 Amp Constant Current @ 15 ms, Ambient 0.138 0.0048
3.5 Amp Constant Current @ 15 ms, Hot 0.1365 0.0040
450 Volt CDU (100 mJ), Cold 0.136 0.0020
450 Volt CDU (100 mJ), Ambient 0.137 0.0043
450 Volt CDU (100 mJ), Hot 0.136 0.0017
TOTAL 0.136 0.0034

Lessons Learned

We learned several lessons from the failure investigation of the Thermite Torch. One lesson that we learned is to listen to the input of all concerned. The personnel responsible for building the torches insisted that they were not willing to proceed into production with processes that were different from those used during the small development phase. If the engineers had listened to the production personnel instead of using their faulty understanding of what was required for the design, sufficient testing would have resolved the difficulties earlier when it was much easier and less expensive to solve them.

Secondly, we should have suspected the design when we found such a wide variation in the performance of the devices. Getting an almost 50% variation in the output is often a good clue that the design is not robust. Further testing of the design to discover the cause of the variation would have led us to discover the source of the variation, and have allowed us to avoid the Qualification failures.

The main lesson is that a tight schedule is no excuse to skip any development testing. It took only a week (of intense effort) to determine the failure mechanism. The same amount of time spent during a proper development phase would have allowed us to enter Qualification with a much better design. Because the proper time was not spent up front, two months were spent ordering and receiving new hardware and reworking the lot to get it ready for Qualification. The old adage "Pay me now, or pay me later" still applies to work in this field.


The authors wish to thank John Adams, Jim Edwards, and David Haas for their assistance in the manufacture, test, and rework of these components. Julie Thomes and Carol Tibbitts monitored the quality aspects of the program, and Dan Knick was instrumental in guiding the program. All of their help was instrumental in determining why the original design failed Qualification, and finding a fix that resulted in a much improved device.


Barry T. Neyer (1994) "A D-Optimality-Based Sensitivity Test," Technometrics, 36, 61-70.