Modernizing TUS Tuning on Heat Treat Furnaces
Maintaining precise temperature uniformity is a cornerstone of pyrometry compliance and part quality in heat treating. Yet, traditional manual tuning of multi-burner furnaces is slow, labor-intensive, and prone to inefficiencies due to nonlinear system responses.
In this Technical Tuesday installment, Ben Witoff, manager of Information Systems and Data Strategy for Fives North American Combustion, Inc, outlines a new linearization-based approach to combustion tuning. This approach offers a data-driven method to shorten uniformity adjustments, improve survey outcomes, and elevate furnace performance to higher AMS2750H classes with greater repeatability and less operator intervention.
This informative piece was first released in Heat Treat Today’s October 2025 Ferrous & NonFerrous Heat Treatments and Mill Processing print edition.
The Manual Uniformity Tuning Model
Heat treat furnaces require precise combustion system tuning to produce high-end parts for aerospace, automotive, and construction industries. Temperature uniformity surveys (TUS) are the accepted industry standard for verifying the quality of these metal processing furnaces. Current standards such as AMS2750H specify a temperature uniformity that must be maintained inside the furnace work zone. End users, like Boeing, GE, and Pratt & Whitney, have mandated these temperature uniformity quality standards for their suppliers who heat treat components for their products.
Today’s furnace and combustion system TUS tuning methods are slow, inefficient, and outdated. These methods require skilled technicians to make precise, manual adjustments to single components in an iterative fashion. Adjustments require this recursive approach because the system of equations governing an industrial furnace’s heat distribution is nonlinear — an adjustment in one part of the furnace system will require tweaking the set up in other area(s).
In the typical case of a multi-burner furnace which has more than one temperature measurement point, the temperature distribution across a measurement array is not directly proportional to the change in a single burner’s firing rate. Due to the system’s nonlinearity, each independent tuning adjustment has incidental, cascading downstream effects on the rest of the system. Every attempt to resolve temperature disparity in one area of the furnace can consequently bring another area out of compliance.
Reinvention of the Tuning Process
Fives North American Combustion, Inc. (FivesNA) has developed a solution that shortens the time of the temperature uniformity tuning process when used before each TUS and optimizes the furnace temperature uniformity. The North American CertiFire panel implements a patented (Robertson and Dzik 2024) temperature mapping algorithm that creates a linear approximation of any furnace’s system of equations, regardless of its geometry or complexity. Once linearized, the temperature distribution can be resolved through simultaneous adjustments.
The temperature mapping algorithm creates a response matrix that correlates changes to the furnace’s heat inputs with changes in the steady-state distribution of heat throughout the furnace’s work zone. A thermocouple array is used to measure the work zone’s three-dimensional temperature distribution while the furnace’s burners are modulated. It is critical to the accuracy of this response matrix that the burner modulations are precise and repeatable. To accomplish this, actuated gas valves are inserted in the gas line to individual burners taking the place of a manually adjusted limiting orifice valve.
Each individual burner modulation has its own characteristic effect on the entire work zone’s temperature distribution. Figure 1 shows two different burner modulations and Figure 2 shows the resulting furnace temperature distribution over the same period. Nine thermocouples were placed on a rack within a furnace in accordance with the AMS2750H standard for this furnace volume and class, with eight thermocouples at each of the cubic work zone’s vertices and one in its center (SAE International 2022, p 44, Table 17).


The firing rate of each burner was increased to a fixed amount for a set number of minutes. The second burner was not adjusted until the work zone’s bulk temperature returned to the baseline average temperature. The two burners noted in Figure 1 were firing in the same plane, several feet from one another. Despite the burners’ close proximity and similar adjustments, their effects on the temperature distribution shown in Figure 2 are uniquely different. Not only does the overall rate of temperature change differ between the curves, but so do the individual thermocouple reactions. Thermocouple 5 (shown in orange), for example, shows the largest change in temperature for the first burner’s modulation, but experiences a much weaker response during the second burner’s modulation.
Mathematical Approach
The linear approximation of the furnace’s system of equations can be written as shown in Figure 3. Where vector T represents the temperatures of q thermocouples, vector B represents the bleed valve modulations of r burners, and response matrix K represents their relationship. By compiling each of these burner modulations and their resulting temperature effects, the furnace’s unique response matrix can be calculated using the formula shown in Figure 3.

Once the response matrix is known, the linearized furnace equation can be reversed. By dividing a vector of thermocouple temperatures T by the response matrix K, the equation yields a vector of automated burner gas valve positions B. In a steady state furnace, starting with a vector ΔT representing the required changes in temperature for each thermocouple to reach the survey temperature, this equation can solve for ΔB, the necessary burner gas valve adjustments to achieve temperature uniformity.
The process of training entails the CertiFire map out the general valve positions needed to bring the furnace close to uniformity. Metaphorically, it is writing the manual of the furnace’s behavior which takes time to develop (a few hours). The tuning operation, which is the next and final step after the training, is like reading the manual and applying the guidelines set by the training. This process takes only a few minutes to dial in the valve positions for improved uniformity, as discussed below in the case study.
Case Study: SIFCO Industries, Inc. – Cleveland, Ohio
SIFCO furnace 8001 is a single zone box furnace with four high velocity burners firing above the load on the left wall of the furnace and four high velocity burners firing through piers below the load on the right side of the furnace. All burners were configured with a cross-connected variable ratio regulator. The combustion air was fixed for fuel-only turndown, and furnace control was achieved through a single impulse air bleed valve, which affected all regulators equally.
Adding the linearizing technology to furnace 8001 required the installation of eight actuated gas valves, replacing the existing manually adjusted limiting orifice gas valves, and a PLC subpanel. This added subpanel controls each motorized actuator independently. To drive the actuators, the subpanel was wired with two inputs from the existing panel: the tuner control variable (CV) over a 4-20mA signal, and the controller set point (SP) over ModbusTCP. The existing control panel was left in-place and is still the primary furnace control interface.
Pyrometry
Furnace 8001 is certified according to the AMS2750H pyrometry standard at the following three temperature SPs: 900°F, 1500°F, and 2100°F. Additionally, the survey process requires first holding the furnace 100°F colder than each SP (800°F, 1400°F, and 2000°F) before increasing the temperature to the desired production SP to prevent overshoot.
According to the AMS2750H standard, the furnace’s internal volume requires nine type-K thermocouples placed at each vertex of the cubic work zone and one at the geometric center for temperature measurements during certification. Furnace 8001 had historically been certified as a Class-III furnace (±15°F). The client’s goal was to reduce the overall temperature span at each of the six SPs to move furnace 8001 to AMS7250 Class-II (±10°F).
Installation Overview
The solution was deployed on furnace 8001 the week of March 31, 2025. Six SPs across four calendar-days (April 4–7) were trained and tuned. The training and tuning algorithms run unattended, so the estimated labor hours to set up and run the process at all six SPs was approximately two hours.
Example Training Experience – 800°F
Training at 800°F ran from 7:52 a.m. to 10:16 a.m. on April 5, 2025, for a total of 2 hours and 24 minutes. The furnace PID tuner was disabled for this training, and each burner’s actuated gas valve was locked in place at its last position. One at a time, each automated burner gas valve was opened by 50% to allow more fuel to flow for three minutes before lowering back to its initial position for 15 minutes. An adjustment amount of 50% was arbitrarily chosen to elicit a strong temperature response.
As each burner’s firing rate was adjusted, a unique temperature characteristic was measured across all 12 thermocouples (SIFCO requires three additional thermocouples for its TUS). The difference in each thermocouple’s temperature rate of change and amplitude is the foundation for the training algorithm’s furnace map.
Example Tuning Experience – 800°F
After the 800°F training was completed, three tuning iterations spaced eight minutes apart at 10:40 a.m., 10:48 a.m., and 10:56 a.m. on April 5, 2025, were conducted. The result of the tuning iterations was a reduction in the temperature span (hottest minus coldest) from 20.2°F to 5.3°F, and a reduction from +4.7°F, -15.5°F to +1.3°F, -4.0°F with respect to the SP, well within the range of an AMS2750H Class-I furnace.

The temperature range, shown in Figure 4, illustrates the initial span on the left half of the chart. A clear reduction in span can be noted at the inception of the first tuning iteration around the 10:40 a.m. mark. Note how not only are the coldest and hottest thermocouples brought in towards center, but the overall spread is also closer centered on SP. The same process was done for five additional SPs yielding the data in Table 1.
| SP [°F] | Operating Time | Temperature Span [°F] | SP Variation [°F] | Effective AMS2750 Class | |||
|---|---|---|---|---|---|---|---|
| Training | Tuning | Initial Span | Tuned Span | Initial SPV | Tuned SPV | ||
| 800 | 2.4h | 15m | 20.2 | 5.3 | +4.7, -15.5 | +1.3, -4.0 | Class-I |
| 900 | – | 10m | 6.3 | 5.7 | +1.9, -4.4 | +1.5, -4.2 | Class-I |
| 1400 | 2h | 1h | 42.9 | 9.7 | +22.7, -20.2 | +4.6, -5.1 | Class-II |
| 1500 | – | 10m | 11.9 | 11.0 | +9.0, -2.9 | +7.2, -3.8 | Class-II |
| 2000 | 2.4h | 25m | 21.3 | 11.5 | +13.9, -7.4** | +3.4, -8.1** | Class-II |
| 2100 | – | 8.5h* | 28.6 | 9.0 | +20.1, -8.5** | +4.8, -4.2** | Class-I |
** Bulk furnace temperature read on average 20°F hotter at 2000°F and 2100°F than the controller SP. Reported SP variations are referenced to the median thermocouple instead of controller SP.
Note that the training was only needed for three temperatures. Since this initial data was captured, SIFCO has conducted three monthly TUS surveys. Prior to each survey, only tuning by the Certifire at the required temperatures was performed. The simplicity, efficiency, and accuracy of the linearizing technology ensures the subsequent client TUS will easily pass and maintain Class II uniformity.
Operating this linearizing technology does not violate furnaces with multiple temperature survey temperatures per AMS2750H because the valve positions are repeatable and maintain the original settings and pressures from the tuning prior to the TUS and will remain in that position until the next tuning (SAE International 2022, Section 3.5.4.1.1k).
Conclusion

The adoption of linearization-based combustion tuning represents a significant step forward for the heat treat industry. SIFCO’s Facilities & Maintenance Manager Nick Klusty reflected on their new capabilities, saying, “We had been struggling to maintain Class III on this 8001 furnace for years. It took days to tune the furnace in preparation for every monthly TUS. Now it literally takes minutes. The North American CertiFire tunes so much better than we could ever achieve by hand. So good that we now have a second furnace available for Class II work, which opens up a new channel for production saving us the time and money of using outside services to heat treat Class II products.”
By replacing manual, iterative adjustments with a data-driven, repeatable process, furnace operators can achieve tighter temperature uniformity in less time, reduce the risk of failed surveys, and expand furnace capabilities to higher AMS2750H classifications. Beyond compliance, this approach enhances process stability and operational efficiency, ensuring that heat treaters are better equipped to meet the increasingly stringent demands of aerospace, automotive, and other critical industries.
References
Robertson, T., B. Witoff, and J. Dzik. 2024. Method and Apparatus for Improving Furnace Temperature Uniformity. U.S. Patent 12,104,788, filed October 1, 2024.
SAE International. 2022. AMS2750H: Pyrometry. Warrendale, PA: SAE International.
About The Author:

Manager, Information Systems and Data Strategy
Fives North American Combustion Inc.
Ben Witoff is the manager of Information Systems and Data Strategy at Fives North American Combustion Inc. After founding the company’s data engineering department in 2019, his work focuses on the development of IIoT-enabled combustion technologies and integrating data connectivity and advanced analytics into industrial processes. A Class of 2023 Heat Treat Today 40 Under 40 honoree, Ben has also been a guest on Heat Treat Radio Episode #77: Algorithmic Combustion Tuning with Justin Dzik and Ben Witoff at Fives.
For more information: Contact Ben Witoff at ben.witoff@fivesgroup.com.

Modernizing TUS Tuning on Heat Treat Furnaces Read More »





























