Fault detection by temperature rise

Friction and lubricant viscosity absorb energy from a system, the symptom of which is localized heating. The heating can be measured by appropriately placed temperature sensors. Obviously there will always be some heating in normal conditions (principally due to viscosity losses), but in abnormal conditions there will be more heating. The trick is to identify the abnormal heating against a background of normal heating and environmental temperature fluctuations.

The most effective way to measure localized heating is to measure temperature gradients (i.e. measure at two locations and use the temperature difference) and to qualify the measurement against machine operating conditions.

Temperature measurement

The sensitivity of the temperature rise method in detecting early wear symptoms is not as good as vibration and oil analysis methods, but because of its relative simplicity and low cost is worth considering. Traditionally good quality Pt100 temperature sensors would be used. Because a gradient is being measured, the sensors need to have good calibration tracking over the environmental temperature range. It is desirable to have better than 0.1°C sensitivity.

Alternatively, thermocouples may be used in a thermopile configuration, the output of which is directly proportional to the temperature gradient, removing the need for sensor matching. Sensitivity can be increased using multiple junctions in a common mounting while still maintaining low impedance for minimal capacitive noise pick-up. The lead wires can be a twisted copper pair and the monitoring device need not have the usual thermocouple support functions such as a reference junction, compensation and linearization.

Note that oil viscosity can increase, causing a temperature rise. While this is not likely to cause destructive wear, it may point to lubrication problems that may cause problems elsewhere - so the effect should not be ignored.

See the Temperature Measurement section of this web site for a practical introduction to the subject.

Thermal imaging

Thermal imaging can provide an alternative way to literally see a problem. However, presently thermal imaging is relatively expensive but prices are coming down. Expect to see permanantly installed smart online thermal image sensors in the next few years.

Qualifying strategies

With the Temperature Rise Method, its is important to qualify the temperature gradient measurement before making a decision to warn or alarm a likely failure. Qualifiers may include:

• gradient exceeds limit for a certain period of time
• gradient exceeds limit after machine operating for more than a certain period of time
• gradient exceeds limit if environment temperature has not changed by more that certain amount for more that a certain period of time
• gradient exceeds limit at a certain period of time after startup
• combination of above

If the operating environment temperature is likely to significantly fluctuate (due to other processes, exposure to weather or direct sun) then great care needs to be taken in placing the temperature sensors to minimize these effects, or alternatively the gradient measurements should also be qualified by environment conditions.

In general, as the sensitivity of the process is increased, the qualification criteria becomes more important. One needs to ensure that the qualification criteria are never so tight as to never or only infrequently become valid.

The alarm levels and the qualifying criteria are generally tuned to the particular application. One would start with a low threshold on the alarm and expect many false alarms until the qualifying criteria has been refined. This process is not too dissimilar to other detection methods.

Conclusions

The Temperature Rise Method is a relatively low cost, easily understood but less sensitive method to detect a potential failure condition.