Downloaded from https://www.everyspec.com There are certain disadvantages to the con- a. If in a lot of 1,000 conventionally designed sumer associated with continuous-sampling. items we desire an assurance of 0.99 of reject- One is that many of these sampling schemes do lng lots with a fraction defective greater than not impose a penalty upon the manufacturer for 0.002, it is necessary to inspect a sample of poor quality. Since no penalty Is imposed, there 900. If two trains in parallel are used, it is is little or no incentive to precontrol the qual- possible to maintain the same level by inspect- ity of product. The only added burden in this Ing a sample of 140.
respect is to Increase the number and/or length b. If we desire a fraction defective less than of qualifying periods. A second disadvantage is 0.0001 in the end item, it may be obtained by that unless the frequency of sampling is more the use of two trains in parallel, each train haty- than 2 percent of the production, there is little mng a fraction defective of less than 0.010. or no chance of detecting spotty quality. This is particularly important where significant de- 5-19. Relationship Between Sampling Plan, fects are concerned. Tolerance Limits. and Safety Factor. (See fig- ure 5-10.) In the case of primary requirements, 5-18. Statistical Aspect of Parallel Design.
which define the physical shape or composition Parallel design is a term applied to a situation of an item, the sampling procedures can, if nec - where more than one route or train is provided essary, be established to require a level of from initiation to result. quality higher than that being currently attained in production. The decision in this case is an Initiation Result economic one, as setting of such standards may ' convenuonai aesign force the producer to screen his product, or to develop a different method of production. One Initiationsnin Result must determine whether the cost of this action Parallel design (three trains) is overbalanced by the value of the result ob- tained.
However, in the case of a secondary Because in conventional design, functioning of requirement the reoults obtained measure not the item is dependent on only one trainfrom the only the effectiveness of the producer, tut also point of initiation to the result, the overall re- the effectiveness of the designer. Most func- liability of the item is limited by the functional tional requirements have this property, and quality of each of the various elements com- considerable care should be taken when deter- prising the train.
These limitations, concerned mining both the design and acceptance para- with quality of functioning, are occasioned by meters that the producer is not required to failure to achieve perfection in the design, meet functional requirements which have little manufacture, and inspection of the component. correlation with the primary requirements, or Since poiectecon is unobtainable, and costly to vice versa. Tho beat approach to this problem approach, there is an obvious limititon on the Is to obtain data from the functioning test, ana- extent to which reliability of the end item can lyze the data statistically, develop tolerance be improved through increasing the quality of the components.
In cases where exceptionally high assurance of functioning or safety is de- I aired, a parallel design may be the only eco- t nomical way to achieve these. If we have an item with two trains in parallel, the probability f of the item's not functioning is squared, as corn- I pared with the probability of not functioning for ( L a conventional design, With three trains in parallel, the probability is cubed, and so on. Thus, a parallel design permits considerable saving in cost of production and inspection. In ASOLUT"E M" cases where this saving outweighs the cost of MINIMUM duplicating certain aspects of the design, the 1.iT 1.5r use of a parallel design ought to be considered.
The following are examples which illustrate the Figure 5-40. Rfteahmos of bsaplq pimp,, advantages. tolerefwt liits, and sWely fctor S-11