AMS data acquisition is highly automated, rapid and produces a large volume of data in a short period of time. Physical changes in the sample and instrument changes over the measurement cycle can result in a degradation of the measurement as can sample in-homogeneity and inter-sample variation (perhaps due to sample preparation). Thus AMS data quality assurance needs to be rapid, and a high level of automation is required. Statistical process control techniques have been developed for exactly such situations and are widely used in the engineering and chemical industries for process control. The techniques are relatively simple, based often on control charts to describe the stability of a process, where stability is interpreted as “nothing unexpected occurs”. A process that is in control exhibits only random variation. Control charts assess statistical control by determining whether the process output falls within statistically calculated control limits. Control charts plot the process data through time (cycles/wheels) and overlay a centre line and control limits on the chart. The centre line represents the expected value of the variable of interest, while the control limits are typically set at 3 sigma above and below the centre line. The control limits are based on how the process has behaved while the centre line is often based on how the process is expected to behave (e.g. for samples with known consensus values or known age material). Measurements falling outside the control limits indicate that “something unexpected occurred”. There are a number of different types of control tests, designed to detect different “out of control” behaviours, ranging from simple outlier detection (one or more points greater than 3 sigma from the centre line), to tests which determine whether there is a run of values below or above the centre line (non random variation), whether the points below/above the centre line are all increasing or decreasing (indicating a trend).
The AMS facility (NEC 5MV pelletron machine) at SUERC and the two radiocarbon laboratories have been working together to assess long-term stability of measurement and C-14 process capability (using FIRI Belfast cellulose and barley mash samples), and the results, based on a two year measurement programme have been used to investigate the usefulness of standard statistical control charts as part of the off-line quality assurance procedures. We will present and summarise the results of this study and recommendations about the general applicability of such process control techniques.
See more of Poster Session I
See more of The 10th International Conference on Accelerator Mass Spectrometry (September 5-10, 2005)