Warning diagnostics for inductively coupled plasma-mass spectrometry

The performance of an inductively coupled plasma mass spectrometer is substantially impacted by changes in radio frequency power, nebulizer gas flow, and sample liquid flow. A simple diagnostic procedure has been developed to automatically ensure that the quality of results is not compromised by system degradation. The diagnosis procedure was studied in two different stages of instrument operation: the normal operating stage and the warm-up stage. For normal operating stage monitoring, a decision tree was derived with the C4.5 induction engine applied to As-based signals allowing instrument performance to be classified as either normal or abnormal. Abnormal signals were further classified as to their probable cause. Signals present in both deionized water and 0.5% nitric acid spectra are sufficient for diagnosis with an accuracy of 96.0% and 94.0%, respectively, allowing the use of 'blank' solutions to monitor instrument performance. For the warm-up stage diagnosis, a standard solution containing Li, In, Ba and Bi, was used. Stepwise regression was applied as the key factor filter while partial least square (PLS) was adapted to be the classifier. In addition to the modeling test, the importance of the intensities and intensity ratios of different masses was studied. By eliminating the standard element signals from the model, the possibility of using water/nitric acid as test solutions was also studied. The overall prediction accuracy for the warm-up stage is approximately 92% using the standard solution or either the water or nitric acid type blank solutions. (C) 2000 Elsevier Science B.V. All rights reserved.