Quality control for on-line analysers

Keeping analysers in service is a critical issue for quality by Alan Reeve

Increased use of on-line process analysers for product certification and process control means that a comprehensive quality control program, such as ISO 9000, is needed for this wide range of specialist instruments.

The following procedures are based on experience with Imperial Oil in Canada, where established routines ensure that only accurate analysers remain in service, particularly on critical duties.

Analysers at Imperial are usually in statistical control before they are used in process control, or for the verification of final product quality. They are also under the protection of a preventative maintenance programme, so that equipment reliability, precision and accuracy are always checked.

Maintenance history of all measuring and testing equipment comes under Quality Control Program (QPC) performance procedures. In addition, an impact assessment is done to consider how failure of the analyser will affect downstream customers.

Back-up testing and monitoring is, of course, increased if this is a critical function.

Analyser validation

Several validation methods are available for analysers, such as:

* Direct sample injection, where a homogeneous sample is injected into the analyser, via a totally enclosed sample system. The analyser result is compared to the validation sample value determined by the inspection laboratory.

* Paired sample, where a sample is removed from a sample point close to the analyser injection point and analysed in the laboratory.

* And an externally purchased standard sample which can be injected into the analyser and compared.

Validation schedules depend on: the type of analysers used; the process involved; the time delay between production and shipment of product; and the cost of off-specification product recovery.

Typically, a minimum of 15 data points are made available over a two week period. Gradual degradation over time can also produce drift that is unacceptable as indicated by the validation sample results.

Should a new analyser be required, the necessary analysers and validation samples are requisitioned during the commissioning process, in order to verify the analyser’s performance.

A quality control schedule typically includes the following.

* Control charts for each analyser. This is the main statistical tool used to monitor the `fit for use’ status of the analyser system. By charting the difference between the previously determined validation sample value and the process analyser result, any change in the analyser system can be detected before a major breakdown.

Typical presentations are histograms, Xbars, moving range/individual charts and, where applicable, range and exponentially weighted moving average charts.

* Data reliability indices, including the ratio of completed validation tests to passed validation tests.

* A test to determine if the bias between the analyser and the inspection lab is statistically significant.

* And, a test to determine if the variability of the analyser has changed statistically between the present study and the previous one.

The main control charts used for analyser quality control are the individual/moving range charts. These charts monitor in a graphical format the long-term performance of the analyser, as well as detecting quickly any change in the analyser performance that cannot be attributed to random noise.

The main advantage of using control charts is that they will indicate when the analyser system requires technical troubleshooting.

* The Author thanks Michaele Sabele of Imperial Oil in Sarnia, Canada.