Downloaded from https://www.everyspec.com limits within which H1e a certain percentage of 5-21. Variables Isopectlo Compared with At- the distribution, determine whether such limits tribute Inspectioa. In attribute inspection each" ane suitable from a functioning standpoint, and, Item is measured, and classified as either good If so, develop acceptance sampling procedures or bed. Often this measure may be the GO, NOT based on the same distribution percentages and G0 type of check. No consideration is given to risks as were used to estimate the tolerance items just outside the limits, or to the extent of limits, variation from Item to item.
Thus, one advan- tage of this method is the ease and simplicity If a safety factor Is incorporsted in the speci- of checking each attribute. Little skill is nec- flied limits, the test becomes one of increased essary, and no calculations are required. severity, and the point of test Is not the real However, sampling by variables requires the point of interest. The problem of developing actual measurement of each item, and the the acceptance requirement in this case is that mathematical computation of measures of cen- of determining the reliability and confidence tral tendency and of dispersion.
The most corn- needed at the specified limit in order to guarai- monly used measures of central tendency andof tee the reliability and confidence desired at the dispersion are the mean (arithmetical average) point of interest (specified limit minus safety and the standard deviation. This gives valuable factor). Data is necessary for resolution of Information about the items under consideration. this problem, for it is necessary to know the It is not necessary to inspect as many items distance between the specified limit and the under the variables plan in order to obtain the limit of Interest, in terms of certain para- same assurance of accepting only good lots.
If meters of the estimated distribution, the item requires a great deal of labor to in- spect It, or if it is damaged during Inspection, 5-i0. Sampling Plas Based on Variables. In- the variables plan would be preferred. spection by variables simply requires each item Certain limitations exist for variables inspec- in the sample to be measured and the reading tion. The distribution must be normal or near- rqoorded, Thus, the exct size of character- ly normaL If it is abnormal, the actual char-, Istic is known, and how much variation occurs acteriatics of the sampling plan will differ from from sample unit to sample unmt is also known.
those upon which the plan was based, and may It can also be determined how much each meas- result in unnecessary acceptance of materiel of urement varies from the specified limits. After inferior quality or rejection of an acceptable the readings are obtained, measures of the av- product. If screening has previously been per- erage and dispersion are computed. To deter- formed, some lots of acceptable quality will be mine the acceptability of each lot, the average rejected by the variables inspection plan. The and, a multiple of the measure of dispersion are reason for this is that in screening a small compared with the specified design limits.
Since!, fraction of the distribution has bben removed, most production processes are normal distri- but the procedure for determining acceptance is bution, or nearly no, one can determine fairly based on the assumption of normality and can- accurately what percentage of the units in the not recognize the effect of the screening opera- lot will exceed any given limit. tion. REFERENCES AND BIBLIOGRAPHY 1. Grant, "Statistical Quality Control." 2. Burr, 'Engineering Statistics and Quality ControL" 5-12