Controlling Temperature
Improves Accuracy.

Precision dimensional control is an important part of manufacturing, yet many companies continue to ignore the single largest cause of precision measurement error: temperature. Typical tests of precision measuring systems fail to address this problem. A short-term capability study on a process or gaging system may lead to the belief that a very acceptable gage R&R, Cp or Cpk can be obtained. However, it may hide the fact that over the long term, thermal effects can cause significant deterioration in performance.

A recent study showed not only the detrimental effect of temperature fluctuation over the long term on an apparently good process control gage, but also the benefit of using electronic temperature compensation to overcome this problem. A short-term study of a gage used to measure mass-produced 18 mm (0.71") diameter spool valves indicated that the process could be controlled with a Cp value of 2.77. That should be pretty good. However, a long-term study, during which ambient temperature increased by just 5° C (9° F), showed a deterioration of the Cp value to 1.56. Moreover, Cpk fell from 2.12 to just 0.35.

Fortunately, the quality manager responsible for system implementation provided for the gage to be equipped with an electronic temperature compensation system. After this was connected, the same long-term test (with temperatures rising by 5° C) produced a Cp of 3.11 and a Cpk of 2.81. Table 1 summarizes the results.

Most shop floor gages are subject to significant thermal variation. They may easily see 5° to 10° C (9° to 18° F) temperature changes between early morning and noon, particularly in summer. Very few mass production plants find it practical or affordable to control temperatures accurately. If a gage run-off is performed in such a factory over a short term, say, half-an-hour to an hour, only small thermal variations will be experienced during the test. However, the tool or gage will be used over much longer periods of time, and then it will be exposed to greater thermal changes. It is rare to test a gage over several hours, or while temperatures change a few degrees, but if the gage will be used under such conditions, it should be tested under those conditions.

The gage in the study was evaluated in this way, and its performance improved by understanding its thermal behavior. In the test, 100 parts were run through the gage three times. Conditions were changed for each run, while data were recorded.

First, the gage's short-term capability was assessed. The 100 parts were sent through a conveyor that fed the automatically controlled gage. The gage controller is equipped with computerized data-collection and statistical process control software. It calculates statistics and displays them in real time. It is intended for use on a production line immediately following a grinding operation to produce outside diameters to micron tolerances.

While attempts were made to control ambient temperatures during this run, there was nevertheless a typically unavoidable morning ambient increase of 0.6° C during the approximately 50 minutes that it took to measure the parts. Figure 1 plots each measurement taken during the run independently; it shows a trend line that indicates a shallow deterioration of about two microns in mean value because the temperature rose slightly during the course of the data gathering. Because the thermal change was not large, Figure 2 shows a histogram with an apparently very acceptable distribution.

Now take a close look at the results of the second run, during which ambient temperature rose by just 5° C, such as might easily be experienced in a morning of production. Figure 3 shows a trend line in which mean diameter varies by eight microns. Note that the mean value decreases. Of course, the reverse might occur during an evening or night shift as temperatures dropped. Because in this case the dimensions are apparently decreasing, this indicates that the gage or gage tooling is expanding at a higher rate than the workpiece.

As further studies determined, the gage was in fact expanding at a very high rate, which is not uncommon. Gages and gage fixtures usually are complex mechanical and electrical assemblies that for several reasons are heavily affected by thermal changes. For example, steel expands at an approximate rate of 6.8 parts per million per degree Fahrenheit, and aluminum at a rate of 13 parts per million per degree Fahrenheit. By comparison, this gage expands and contracts at the rate of 33.3 parts per million per degree Fahrenheit. Because the gage fixture expands as temperature increases, the measuring probes must extend further so as to contact the workpiece, thus giving the impression that the part is smaller.

In a practical application, this could lead to acceptance of oversized parts. In a system in which machine compensation (automatic feedback to a machining operation) was being used, the gage would actually instruct the machine tool to make oversized parts. On most shop floors, temperatures vary continuously. As temperatures change, so will measured values of critical dimensions. Figure 4 shows the effect that the 5° C drift had on the distribution of measured diameters during the second run. Clearly, the wider distribution of measured data points led to a considerable reduction in calculated capability (Cp), and the trend in the direction of decreasing mean size caused a major deterioration of Cpk value (capability centered around nominal size).

For the third and final run, a GageComp temperature compensation system from Albion Devices Inc. was added to the gaging system. The custom-designed industrial sensors monitored temperatures of workpieces, the setting master and the gage as parts were measured. GageComp then calculated a correction based on predetermined correction coefficients, and electronically sent this correction in real time to the gage. The computerized gage mixed the correction with the measured dimension to produce a temperature-corrected measurement of each part. Effectively, the gage now displayed the dimension that would have been obtained if each measurement had been taken while parts, master and gage were held at the International Reference Temperature of 20° C (68° F).

Compare figures 2, 4 and 6 (runs 1, 2 and 3, respectively), and you will see that original gage capability (run No. 1, Figure 2) is severely compromised when temperature variations influence measurements (Figure 4). By applying temperature compensation while temperatures changed in run No. 3 (Figure 6), original gage capability (run No. 1) is restored. In fact, it is even improved.

As Figure 5 shows, the mean dimensional measurement remains flat when temperature compensation is applied. After smoothing out inherent gage R&R variation, average measurements were virtually unaffected by the change in ambient temperature. Further, as seen in Table 1, Figure 6 shows an improvement in both Cp and Cpk over run No. 2 (in which ambient temperature increased by 5° C) and even run No. 1, in which temperatures increased by just 0.6° C. This last observation demonstrates that process improvement can be achieved using temperature compensation even in environments where temperatures are held stable to within a degree or two, which is the best that can be held in large plant areas.

Electronic temperature compensation systems are widely used in a variety of applications. They can interface with just about any electronic gage, but think of them as separate, distinct systems from the mechanical gaging assembly. When discussing this subject with gage suppliers, keep this in mind, and differentiate responsibilities among vendors of the different system components. It can be useful to establish criteria for gage performance first at stable temperature, then under typical shop floor thermal variations. By separating these requirements, specific responsibilities can be established for performance under varying conditions.

A relatively small investment of time and money in temperature compensation can pay huge dividends. Increased process capability is well-known for yielding cost savings and, more importantly, providing a competitive edge. Mass producers of discrete parts invariably are under pressure to produce, as are production and inspection equipment suppliers. Understandably, they may want to avoid the time-consuming task of a long-term study.

However, as the above results clearly reveal, a short-term capability study is not necessarily a true test. Customers who rely on such studies to certify the quality of a supplier's process should beware. Companies that take advantage of this opportunity to improve process capability will certainly become more competitive in their marketplace.

About the author

Paul Sagar is president and co-founder of Albion Devices Inc. of Solana Beach, California, ( www.albiondevices.com ) which develops proprietary technology for thermally analyzing and automatically correcting precision gages. Albion's systems are used at major automobile, aerospace, bearing and printing companies, and major railroads. E-mail him at psagar@qualitydigest.com .