Market Topics

How Good Is Your Data?


A significant tool of reliability centered maintenance is the use of Predictive Maintenance (PdM) to monitor machinery condition. The monitoring of in-service lubricants is a vital PdM technology that has received enhanced interest over the last several years. Strategies include the use of in-house mini-labs, full-service corporate laboratories, and the use of external test service providers.

The weakness in much in-service lubricant testing performed, however, is the unknown accuracy of the data produced. This uncertainty in accuracy stems from the lack of consensus standards and use of unsupported modifications to existing standards.

The reasons for this gap include customers not demanding data of known accuracy; the use of proprietary test methods or test method revisions by the testing service providers; and standards organizations longtime emphasis to support new product testing and development.

As a standards development body, ASTM International has a very active program devoted to the testing of lubricants. Its standards committees are made of volunteers who meet on a regular basis to address the standardization needs that are brought forward. Other organizations such as The Energy Institute in Britain, Germanys DIN and the International Standards Organization also provide standards related to the testing of lubricants.

The significant majority of ASTMs petroleum-related standards are written to address testing of new lubricants. These standards aid in the manufacture and marketing of petroleum products. The benefit to the end-use consumer is in known physical properties of the product being purchased rather than in-service lubricant monitoring. ASTM standard test methods include text that describes how to perform a test, a description of the apparatus used, and a section associated with the accuracy and precision of the test method.

This last – data accuracy and precision of test results – is the vital function of a standard test method. The importance of this is in knowing the expected variance of data. Stated otherwise, its how accurate the result will be when the steps of the standard are carefully followed. Equally important is how the result will compare, or if it even will compare, when the sample is tested by another laboratory.

Data accuracy or precision is determined in most standard-setting bodies by organizing a group of committee participants willing to follow the steps of the test method verbatim and apply them to a common group of samples. This process is known as a round-robin. ASTM test methods contain precision and bias statements that are based upon such round-robins. A research report approved by the sponsoring committee and filed with ASTM upon conclusion of the round-robin discusses the methodology used, laboratory data generated, overview of the statistical treatment of data, and the resulting precision statement for method repeatability and reproducibility.

The rigor in determining the relative goodness of the test results is a key difference between test results based upon standard test methods and ones based upon in-house methods. These differences can readily affect the pocketbook. Varying the approach to testing varies the steps needed to obtain data, and obtaining test results from modified methods in lieu of standard ones can be expected to reduce cost.

Reasons often given for modifying a standard include: The steps are really not that important, the data obtained through our modification is good enough, there isnt a standard available but the one referenced is the closest one to what we do, or simply, I can do it cheaper my way. Modifications to the standard can and likely do mean that portions of the standard were omitted or revised by the laboratory. How accurate the data needs to be to be good enough is a valid and appropriate discussion point, but has not been addressed through the standardization process.

For condition monitoring purposes, the cost incurred in obtaining the data to meet ASTM precision coupled with the expected use of the data needs to be a match. In many cases, however, there isnt a good match of ASTM standards and their cost versus that of service-provider laboratories in-house methods. Many in-house methods will provide very acceptable test data. The explanations for using data of unknown accuracy/goodness, while seemingly reasonable, still leave the customer requiring the data in the dark.

The bottom line is that the end user of data produced outside of a consensus standard has no assurance of the relative data goodness of the test results. When a test is not anchored to a standard, it produces data of unknown quality. This doesnt mean that the data is unusable or unreasonable; rather that its intended use should be carefully evaluated and considered. An example of this is in the measurement of viscosity.

Viscosity is the most common and perhaps most important test performed on an oil sample. ASTM D 445, Test Method for Viscosity, is designed to be used on transparent and opaque fluids. So far this standard would seem to be a match for in-service or used oils as some used oils are transparent while others are opaque. While this standard can certainly be applied for used oil testing, it does not contain a section that discusses test precision or accuracy for used oils. Its use or reference by any laboratory for used oil testing would certainly be a modification of this standard, resulting in data of unknown quality.

This particular problem is being addressed by ASTM, and a precision statement for used oils is expected to be included in the next version of the test method.

On the other hand, one might argue that the precision of D 445 (0.26 percent) is much tighter than needed for condition monitoring purposes. For example, an ISO 68 viscosity grade oil can be manufactured to fall anywhere from 61.2 to 74.8 mm2/s at 40 degrees C, and still be classified as an ISO 68 oil. This is a range of plus or minus 10 percent, which is significantly greater than the 0.26 percent variance of D 445. So the argument could go, Why should this level of accuracy matter? Some instruments available on the open market claim viscosity repeatability of plus or minus 5 percent. This variance may be acceptable for many in-service applications but likely would be unacceptable when monitoring a critical application such as turbine oil. With a variation in the range of 0.26 to 10 percent for viscosity, the possibility exists for great variation of data precision in the testing community.

In December 2001, ASTM formed a new subcommittee – Coordinating Subcommittee 96 – to write new standards and coordinate revisions to existing ones for the benefit of the in-service lubricant condition monitoring test industry and its customers. This new group is made of volunteers from commercial laboratories, instrument manufacturers, oil companies and end users of the data. The goal is to cover the range of testing performed by the in-service lubricant testing community. The expected primary benefit to the consumer will be data of known quality.

Two types of standards are being written by Cordinating Subcommittee 96. The first encompasses instrumentation that has wide current industrial usage but has not been standardized. There are many instruments in use that fit this category, and viscosity measurement is an early example of this subcommittees success. ASTM D 445 is now being edited to include used oil as an application, and new standards are at various levels of development to address the more than 7,000 viscosity instruments currently active in the field.

The second type of standard for the used oil community will be based upon a laboratory demonstrating a quality process that will include audits, written procedures, training and trending of instrument performance. These performance-based standards will umbrella the laboratories and allow use of their own methods. The laboratories that use this type of standard will be required to tie their data to a reference material to demonstrate accuracy, and then would report their results to their customers in graded degrees of precision (e.g. level 3-fair, level 2-better, level 1-best). This standard is not intended to place ASTM in an accreditation role; rather this function will be an issue to be resolved between customer and service provider.

The end user of test data has a vital role in influencing the quality of data produced by condition monitoring service laboratories. When no expectations are given, the quality of data is entirely in the hands of the service providers. The customer of the data can and should demand data of known quality. Where standards are not available, a demand on behalf of either the service providers or customers can be made to standards organizations to develop used lubricant monitoring standards that provide a measure of data quality.

The end user should be heard in the marketplace as well as within organizations such as ASTM. ASTM D02 and Coordinating Subcommittee 96 will welcome all those interested in participating in this process.

Related Topics

Market Topics