Getting the Measure of Temperature

Jonathan Pearce explains how new techniques can improve process efficiency

KNOWLEDGE of temperature is essential for measuring and controlling almost every technological process. In recent years, substantial advances have been made in techniques to measure temperature across a wide range of disciplines applicable to chemical engineering.

Measuring the temperature of a substance directly is really hard because it is a measure of the average energy of the constituent particles. A so-called “primary thermometer” is needed, which measures some parameter, such as the speed of sound in a gas or the Johnson noise voltage across a resistor (more on this later), which can be related to temperature through well-understood physics. Unfortunately, these methods are complicated, time-consuming, and not at all practical.

Measurement traceability

Most of us, then, are forced to use more practical sensors such as resistance thermometers or thermocouples. These rely on the temperature dependence of some property of the thermometer, such as resistivity or the thermoelectric effect. Since these properties are too difficult to relate to temperature directly – the physics is just too hard – a calibration is required. Happily, there is a measurement infrastructure which makes this possible, and at the heart of it is the International Temperature Scale of 1990 (ITS-90)¹.

This infrastructure is maintained by National Measurement Institutes (NMIs), which in the UK means the National Physical Laboratory (NPL). ITS-90 allows temperature measurements to be traced back to the SI unit of temperature, the kelvin. The scale has two components: firstly it defines a set of specific temperature values, or fixed points, which users can realise, and which have been established via international consensus. These are typically highly-reproducible values like the freezing or melting temperatures of pure materials. Secondly it provides a recipe for interpolating between those fixed points². A thermometer calibrated in accordance with ITS-90, and whose calibration path can be traced back to an NMI, permits traceable temperature measurements. This assures the user that the temperature measured is one that can be reproduced, both in time (a future measurement of the same system will be comparable to the earlier measurement) and in space (measurements across different plants around the globe will be comparable), because it can be traced back to an internationally-validated standard.

However, a key drawback of this approach is that when the thermometer is placed in a harsh environment – for example, a furnace used to heat treat aerospace components at 1,300°C, a steel manufacturing process at 1,600°C, or the coolant circuit of a nuclear reactor at only a couple of hundred degrees but subject to high doses of ionising radiation – the changes to its materials will cause a progressive, unknown change in the calibration – a so-called “calibration drift”. This article will explore some up-and-coming solutions to overcome this difficulty.

The reproducibility crisis

Imagine if you were presented with two temperature measurements taken by two different people – for the same process – yet they differ by 30%. Which one can you trust? Now extrapolate that to all the temperature measurements made in your process.

Recently this reproducibility crisis has been discussed at some length in the scientific literature³ and there is now growing awareness in industry about how metrology, the science of measurement, can help⁴. By relating measurements to reference standards which have been internationally agreed with known uncertainty, their equivalence between different practitioners can be assured. In the case of temperature, using ITS-90 with appropriate traceable calibrations (look for calibration labs with ISO 17025 accreditation⁵) provides that assurance. In turn, this results in improved efficiency and reduced wastage – for example through the ability to operate within a tighter process envelope – and the ability to operate across multiple sites with the assurance that all processes are equivalent (see Figure 1).

Calibration drift

Calibration drift arises when the sensor is changed by factors such as chemical contamination (eg from hydrocarbons in the exhaust of a jet engine, or silicon processing and semiconductor manufacturing), or high temperature damage (eg during casting of single crystal nickel alloy gas turbine components, growing large crystals of quartz for the manufacture of optical fibres, or cement manufacturing). This in turn changes the relationship between the sensor output (such as resistance or voltage) and temperature, ie a change to the calibration. The practical effect of this is a progressive loss of knowledge of the temperature, as the error becomes progressively larger. This is particularly problematic for thermocouples, which are typically used in much harsher environments than other thermometer types.

Dual-wall thermocouples

For temperatures up to about 1,200°C, commonly-used thermocouples include the mineral insulated, metal-sheathed (MI) Type K and Type N thermocouples. The swaged, flexible format of these types makes it simple to incorporate them into processes, but they are subject to calibration drift, especially at high temperature because the thermocouple wires can be contaminated by the outer sheath.

Recently, Cambridge University developed a new variation of this type of thermocouple, which uses a dual outer wall consisting of two layers⁶. The outer layer is of the usual material (typically stainless steel), while the inner layer is of a material which helps to reduce the effect of contamination of the thermocouple wires from the outer wall at high temperatures, where diffusion of the thermocouple materials causes cross-contamination throughout the unit. The stability of this dual-wall thermocouple has been shown to be 5–10 times better than that of the conventional format. These thermocouples have been licensed to UK thermocouple manufacturer CCPI Europe, and will soon be commercially available.

Optimised Pt-Rh thermocouples

Thermocouples based on platinum and rhodium (Pt-Rh) are the workhorse of thermometry above about 1,200°C where other types degrade too quickly. However, the commercially-available Types R, S and B all contain wires whose alloy compositions are quirks of history. In 1885 an ad hoc thermocouple consisting of Pt-10%Rh versus Pt wires (now referred to as Type S) was first investigated by Le Chatelier, who needed a stable thermocouple to reliably control cement manufacturing. Use of this type of thermocouple snowballed, and by 1900 it had become popular. However, it was eventually discovered that in Britain the wires were inadvertently contaminated with iron, which came from the rhodium, and which changed the calibration significantly, as well as aggravating calibration drift. By the time this was discovered, use of the Type S thermocouple had become widespread, so a new iron-free thermocouple, the Type R, was prepared using Pt-13%Rh to match the existing temperature versus voltage characteristics of the iron-bearing Pt-10%Rh alloy. This expedient continues to this day. Similarly, the Pt-30%Rh versus Pt-6%Rh thermocouple (Type B) was first made available in Germany by Degussa in 1953 to access higher temperatures, but it is not known (at least not to me) how they arrived at this formulation.

So there is an opportunity to optimise the composition of Pt-Rh thermocouples, and at NPL we’ve made a concerted effort to determine wire compositions which give the lowest calibration drift. By testing a large number of wires of different composition using a new generation of high accuracy high temperature calibration techniques², a new optimal pair of wires has been identified, yielding the Pt-40%Rh versus Pt-6%Rh thermocouple. Although new, this thermocouple is readily constructed using commercially-available wires, and work is ongoing at several NMIs to produce a standardised relationship between voltage and temperature, which will facilitate off-the-shelf use and accurate calibration.