00043 D1700e75ed9e1e1afbfb3d368d6e171

42

Molnau

detection. This value is not reached until the 30th defect. Again, sińce a complete nuli scenario is also used, all readings 1 - 30 are disregarded and extra runs madę to compensate for this. Similar to the sudden shift case, all detection classifications received are termed correct detections, while all false alarm classifications are termed rnisses.

Results

The OPS5 production system produces output reports to summarize the results of input test data. As discussed in previous sections, the system reads three required data elements when an out-of-control signal is received. The system takes the data values, and makes a determination whether the signal is a correct detection of a false alarm.

Figurę 8 shows a sample output of a detection signal. In this case, the system correctly diagnosed the signal as a false alarm. The justification for this judgement was that rule 1 fired--the additional CUSUM metric reading was below the 0.5 level. Every incoming signal to be evaluated is processed in a similar manner.

Figurę 9 displays the finał run results for one of the upcoming test scenarios—the sudden shift case. The report summarizes the system's performance for a set of runs, and compares the performance with that of the leadtest.

CUSUM SIGNALED AN OUT-OF-CONTROL AT THE U 42 DATAPOINT. THE TEST YALUES ARE:

Data

TEST

TO YALUE

T5YALUE

U OF F1RES

CUSUM

CUSUM2

SPRT

EWMA

GCC

1.054419e+000 8.435350e-001 9.967400e-001 1.013880e-001 1.543200e+001

2.876030e-001 2.300820e-001 3.685089e-001 1.849188e-001 2.548341e+001

0

0

0

0

THE SYSTEM DETERMINES THIS TO BE A MISFIRE RULE1-CUSUM(5) YALUE LESS THAN 0.5

Diagnosis

IS THIS DETERMINATION OK? Y/N y

THE CURRENT SYSTEM SUCCESS RATĘ IS 9.500000e-001

Figurę 8. Sample Detection Signal Recommendation.