Temperature measurement is a critical parameter in many industrial processes. As a result, these processes generally use installed thermometers or temperature transducers. Proper calibration of these installed systems has always presented a challenge. Traditionally, simulators are used to provide in-situ calibration of the electronic portions of these systems. Although this is a valid technique for the calibration of the electronics, considerable inaccuracies are introduced by the sensor itself.

The metrologist is often presented with complex measurement problems in the laboratory. If the technology is available in the form of standards and instrumentation, solutions to these problems are generally straightforward. We systematically apply our metrological tools to the situation, perform analysis, make decisions, and arrive at a solution.

Our tool bag includes mathematical models, error budgets, statistical methods, error propagation equations, and many other tools. If we apply these tools properly, with consideration to the financial aspects of the situation, our solution is usually cost effective and workable. Process calibration problems can be dealt with using the same techniques. The challenge for the metrologist is to be flexible when analyzing process related calibration problems. To be effective, we must realize that we are consultants to the process engineers and technicians. We are not running the process. The solution must be appropriate for the situation and workable. If possible, the solution should provide for calibrations on site, right there at the process. Down time in the factory is usually extremely costly and must be minimized.

Many of these industrial processes require temperature measurement of one kind or another. This statement holds true for both manufacturing and testing processes. In fact, it has been stated that temperature is the most commonly measured process parameter in industry today. In many cases the dependability of these measurements has been either totally ignored or taken for granted. Traditionally, the solution has been to calibrate the measurement instruments very carefully while neglecting the sensors themselves. With modern instrumentation, the electronics is a very small component in the total error of the system. Much of the error comes from the temperature sensor or the signal cable.

Why calibrate process measurement systems?

There is a current trend in manufacturing and testing to better define these processes from a measurement standpoint and apply some of the same rigorous procedures that we are familiar with in metrology. There are several reasons for this trend, and they can be grouped into three categories:

1. cost reduction
2. competitiveness
3. accreditation

Cost Reduction

Industrial processes consume energy and energy is expensive. Obviously, one direct way to reduce the cost of manufacturing an item is to use the energy required in the manufacturing process more efficiently. One example of proper process management is accurate regulation of industrial dryers. If the dryer is hotter than required, energy is wasted. If the dryer is on too long for the actual temperature and humidity conditions, again, energy is wasted. Industrial dryers consume extremely large quantities of energy.

Tens of thousands of dollars per year can be saved directly from reduced energy consumption if drying processes are managed properly. A second example of cost reduction is to reduce the waste product by better control of temperature dependent process. High technology ceramic components require precise drying and/or bake-out profiles. If they are exposed to incorrect temperatures during the drying process, component strength is compromised. These components are generally lot tested for strength and poor performing lots are scrapped. High technology ceramics are expensive, and waste can pose a significant cost problem.

Competitiveness

Along with the cost, the quality of a product or service affects its competitiveness. Proper process management impacts this aspect as well. Obviously, reduced production cost for a product allows the manufacturer increased flexibility in the marketplace. We have touched on a couple of examples of cost reduction in the previous section. Here we will focus on quality. The quality of the ceramic components that we considered earlier is dependent upon temperature at several stages in the manufacturing process.

Slurry temperature, drying temperature, and firing temperature all affect the quality of the component. Errors in the temperature at these stages can result in poor and inconsistent quality. Another example of how temperature affects quality is that of a thermistor (temperature sensitive semiconductor) manufacturing process. Thermistor response is mainly dependent upon the dopants in the mixture that composes the semiconductor itself. However, as with ceramic components, the firing temperature effects it also. Additionally, once the material is manufactured and shaped, the resulting thermistors have to be tolerance sorted. This sorting process is used to verify the temperature vs. resistance relationship. The accuracy of the temperature determination obviously impacts the quality of the component.

Accreditation

Manufacturers who wish to become accredited are being compelled to calibrate process parameter measurement systems. In many situations, the systems have been aligned or calibrated in the past, perhaps inconsistently, and the manufacturer simply has to establish procedures and follow them. In some cases, however, the systems have never been properly calibrated and there are no established requirements or a baseline for performance. In these cases, the manufacturer may have to consult with experts in the discipline for assistance.

For example, sheet plastic extrusion machines are temperature controlled along the entire length of the machine. The effect of temperature on the quality of the product is only approximately known. Most manufacturers of sheet plastic do not calibrate the temperature sensors in the machine at all and do not have an infrastructure in place to accommodate calibration. To comply with ISO-9001 guidelines, a calibration process may have to be established and followed. The problem in this example is that the manufacturer does not know what his requirements should be. You have to start from scratch and develop it.

Process system examples

Example 1: Nuclear Power Plant Cooling Water

The first example is a cost reduction problem at a nuclear power plant. Nuclear power plants often use ocean water in the secondary cooling system. Federal regulations limit the thermal load that the plant can put on the outlet site (the ocean in this case). To maximize the efficiency of the plant, management wants to run it as close to this limit as practical. If the limit is exceeded, very large fines can result. The outlet water temperature is measured on a periodic basis, but the models are based on engineering tests run during a scheduled refueling outage. These tests consist of simple inlet and outlet water temperature measurements followed by application of mathematical analysis. Because the engineering tests are run during an outage, they must be run at a small percentage of full output (≈20%) and the results extrapolated to full operating load. The equations are linear, so the engineers believe that extrapolation of the results is appropriate. This extrapolation however, places very tight limits on the allowable temperature error during the tests because the error is multiplied by the extrapolation.

The tests were to be run between 15 and 60 °C (≈ 60 and 140 °F) with an accuracy of 0.03 °C (≈ 0.05 °F). This accuracy limit would provide results required after extrapolation. During actual operation, the water temperature is monitored with installed 3 wire RTDs connected to 0.05% accuracy 3 wire resistance bridges. The RTDs are calibrated in the laboratory to an accuracy of 0.03 °C over the entire temperature range. The lead wires for the RTDs are from 10 to 30 meters depending on the location of the RTD. The RTD lead resistance was high in comparison to the resistance of the element because of the length of these lead wires. Before doing any actual evaluation, the team knew that the resistance bridge was not accurate enough for their purpose. They felt that replacing it with an up-to-date instrument would solve the problem.

Table 1: RTD System Accuracy

The team was correct in assuming that the resistance bridge would not provide acceptable results. Even with a precision readout with an accuracy of 0.005 °C, the total error would still exceed the limit by 0.075 °C. Another solution had to be found. Reviewing the requirements of the system would allow us to come to a workable solution.

The requirements were:

• High accuracy
• High stability over time
• Modest temperature range
• Reasonable cost

Additionally, the team wanted to control the test with a computer and use extra sensors for redundancy. It was decided to use custom 4 wire 6 k? thermistors with a scanning readout. The readout would be portable allowing for location at the test site thereby providing a significant reduction in lead length. With 4 wire 6k? thermistors and reduced lead length, the error due to the leads is negligible. The calibration laboratory was consulted by the engineering team and the lab personnel felt that it would not be difficult to calibrate the thermistors and the readout to the uncertainties required. Thermistors were not available to fit the existing thermowells, so new thermowells were fitted during the refueling outage. The final uncertainty values are shown in the table below.

Table 2: Thermistor System Accuracy

In this example, the solution to the problem was an application of laboratory instruments and technique to field use. The equipment was purchased and calibrated, and the tests were completed on schedule and under budget. The engineering team was extremely pleased with the results. Both the engineers and the calibration laboratory received kudos for the success of the project.

Example 2: Thermistor Manufacturing Process

The second example is a quality/competitiveness issue at a thermistor manufacturing facility. In this situation, a thermistor manufacturer was having difficulty getting agreement between the manufacturing process and the quality control (inspection) process. Thermistors that were manufactured and sorted very carefully were being rejected by quality control. This resulted in repeated testing, waste product, and shipping delays.

This manufacturer is very conscientious and will not ship questionable or substandard components. At the time that this problem was presented, the manufacturer was in fact meeting its deadlines. However, they were concerned that if they did not solve the problem soon it would cut into the profitability of the product. This would raise the cost of the product and cost sales, perhaps even damage the company’s excellent reputation.

The accuracy required for this application was extremely high for a manufacturing environment. They wanted total system error of less than 0.010 °C between 0 and 150 °C both on the factory floor and in the QA laboratory. The first order of business was to determine if the error was due to the production testing equipment, the inspection testing equipment, or both.

The engineers on the project were not confident that either area was meeting the tight accuracy requirement. Several high accuracy thermometers were procured, and it was determined that the production baths were not sufficiently stable or uniform to accomplish the purpose. To provide the accuracy required, laboratory baths, readouts, and sensors (SPRTs in this example) were adapted for use on the factory floor and put into operation. Additional SPRTs and baths will be installed in the QA laboratory and SPRTs will be used to ensure standardization between the two areas. Refer to the table below for the accuracy budget.

Table 3: SPRT System Accuracy

This example was similar to the previous one in that the accuracy requirements were very high. The only solution was the application of laboratory instruments and technique for use in a non-traditional setting. When accuracy requirements are so demanding, all sources of error must be evaluated very critically. As in the previous example, the attention to detail paid off and the solution was implemented with excellent results.

Example 3: Ceramic Dryers and Firing Furnace

This example considers an accreditation issue at a high technology ceramics manufacturer. This example represents a more typical situation where the instruments involved were on a calibration and recall system, but the actual sensors were not. This manufacturer was attempting to achieve ISO-9001 certification. The auditor explained the need to include the sensors as well as the instruments in the calibration and recall process. The project manager was under the impression that if “certified” thermocouple wire and DIN class A or B RTDs were used, calibration was not necessary. The engineers were also unsure of the actual accuracy the process had to achieve to produce high quality product with consistent performance. There were two processes that they were particularly concerned about:

1. The drying process
2. The firing process

In the drying process, the components were subjected to relatively low temperatures in a chamber while the humidity was being monitored. The chamber sensors were two wire DIN 0.5 class A RTDs. A programmable chart recorder with an accuracy of 0.25% of span + 0.1°C was hard wired to the sensors and used to monitor and control the temperature of the chamber. The connecting leads between the probes and the chart recorder were approximately 2 meters in length (about 1.2?). On a routine schedule, the chart recorder was calibrated by means of a simulator instrument connected in place of the RTDs.

When the actual temperature of the chamber was spot checked, it was found to be several degrees off of the chart recorder indication. Analysis of the error sources uncovered that the error due to the uncompensated RTD leads was by far the largest contribution. In this example, even if the chart recorder is calibrated perfectly, the error due to the rest of the system would remain large. An additional source of that has thus far been ignored is the DIN conformance of the RTD. Although the probe in our example is stated to conform to the DIN curve within 0.5 of the class A specification, it did not, and they seldom do. Generally, probe manufacturers state that the element used in the probe meets spec, but once it is assembled into a probe, all bets are off. Usually, this results in a very large error at 0°C. In this case, the error was about 0.75 ?, or 1.92 °C.

Table 4: RTD System Accuracy

Clearly, the largest sources of error were due to the RTD nonconformity and lead resistance. The manufacturer could have eliminated these sources of error by replacing the 2 wire RTD with a 4 wire RTD and performing a custom calibration on the RTD in the calibration laboratory. The major problem with this solution was that the chart recorder was configured for 2 wire RTDs only. Use of a 4 wire RTD would require the purchase of new chart recorders. That was not an option. Fortunately, the chart recorder was equipped with zero and span adjustments which had enough range to accommodate the rather large errors due to the RTD. The solution chosen was to use an external standard PRT and recording (data logging) readout to calibrate the chart recorder and RTD as a system.

The standard PRT was in close proximity to the installed RTD, and the chamber was cycled between minimum and maximum values. Adjustments were made to the chart recorder to achieve agreement with the standard. Subsequent calibration would be performed in a similar manner except a third measurement would be made at approximately 50% of the span. This would provide for as found data adjustments would be made only if necessary.

In the firing process, the components were subjected to very high temperatures in a furnace for a predetermined period. The furnace sensors were type S thermocouples. A programmable ramp and soak controller with an accuracy of 0.2% of span + 0.1°C was hard wired directly to the thermocouples and used monitor and control the temperature of the furnace. As in the previous example, the controller was calibrated on a routine basis by means of a simulator instrument connected in place of the thermocouple.

Because of their experience with the dryers, the project manager knew that the thermocouple required calibration (verification). In this case, the solution chosen was to remove the thermocouple probe from the furnace and, while leaving it connected to the controller, compare it to a standard in a portable drywell type furnace. Purists would suggest that thermocouples cannot be calibrated, and they are correct.

This solution employs a verification of a process system. The controller is adjusted to correct for changes in the thermocouple. When this calibration process was first implemented, several UUT thermocouples had to be replaced because they had drifted so much that the offset necessary to bring them into conformance was outside the range of the controllers. It was also noted that most of the error present was due to the thermocouples rather than the controllers.

These examples were traditional process situations where the accuracy of the sensor was assumed. The instruments were calibrated on a routine basis, but large errors were still present. Evaluation of the error sources showed that the largest contribution came from either the sensor itself or the connection to the process instruments. In both cases, the solution was not difficult to implement once the errors were identified. Currently, the systems are on a routine schedule and the accreditation process is proceeding.

Example 4: Plastic Extrusion Machine

The final example considers a plastic manufacturer attempting to achieve accreditation. At issue was a sheet plastic extrusion machine. Unlike the previous examples, this manufacturer did not have anything calibrated on a routine basis. The only time that the instruments were calibrated was when they malfunctioned and required repair. The thermocouples in the machine were replaced if they went open or shorted. Most of the thermocouples were original to the installation of the machine. No documentation existed and no one was sure if the thermocouples had ever been calibrated or even what type they were.

Because they were attempting to achieve ISO-9001 certification, the production management consulted with a sister organization who had a calibration infrastructure in place. They were unfamiliar with temperature calibration and had no idea how to proceed or what the accuracy of the temperature measurements ought to be. Since calibration had never been performed, and the plastic seemed to be of acceptable quality, it was concluded that the accuracy requirements were not extreme. The production engineer had a desire to keep the calibration process simple and as close to the production floor as possible.

The extrusion machines are very long machines with several temperature zones ranging from 50 to 300 °C. There are from 20 to 30 type E standard limit thermocouples installed in each machine. The thermocouples are 20 to 22 mm in length and 4 to 5 mm in diameter. They are connected to the controllers with extension cables up to 20 meters in length with one transition at a junction block. The reference junction compensation is performed at the controller rather than at the junction block. The controllers are rated at an accuracy of 0.5% + 0.25 °C. The error due the extension wire is of unknown magnitude. The errors are summarized in the table below for a value of 200 °C.

Table 5: Thermocouple System Accuracy

In keeping with the production engineering staffs wishes, a simple on-site solution was sought. A two-step procedure was decided upon. First, the controllers would be calibrated on site with a simulator instrument. This would keep the electronics in line. Second, the thermocouples would be reconnected to the controllers and the system would be calibrated in a drywell calibrator. This should calibrate out errors due to the extension wire. If the errors are larger than the individual components suggest they should be, the extension wires would be suspect.

Recommendations

These examples of industrial processes demonstrate the increasing requirement for calibration in the industrial environment. As we attempt to squeeze more and more out of these systems, their performance becomes increasingly critical.

If we keep in mind the principles that drive metrological work, in conjunction with the purpose of the processes you may be called on to assist with, metrologists will be able to contribute in a meaningful way.

First, listen very carefully to the engineers, managers, and technicians involved. They are the ones responsible for the process and must follow through after we leave the project.

Second, dissect the process with the goal of uncovering every source of error present. In the examples at the nuclear power plant and ceramic process, the largest sources of error were the temperature sensors themselves, components that the engineers did not even consider in their analysis.

Third, provide real-world, cost-effective solutions. The most accurate, aesthetically perfect solution is of no use if it is impossible or extremely impractical to implement. However, do not be afraid to recommend a solution that involves adaptations of high accuracy laboratory procedures to the factory floor if that is what is necessary. The nuclear power plant and thermistor manufacturer required such solutions. Both were implemented with excellent results.

Finally, understand that some of the principles that we use every day might be foreign to even very good, highly experienced engineers, technicians, and managers. The whole idea of calibration was a new one to the engineers and managers involved in the plastic extrusion process.