The reliability growth curve exhibits approximately a straight line function when plotted on log-log graph paper. A measure of reliability, such as MTBF, is projected as a function of total test time. For program planning, the test time required to achieve a given reliability threshold may be estimated after selection of an initial starting point and establishment of a growth slope.
The Starting Point. An arbitrary selection of a starting point for reliability growth testing is not recommended, although traditionally values from 10% to 30% of the predicted MTBF have been used. Techniques for determining a more realistic value are given in MIL-HDBK 189 and are based on design analysis and experience on similar systems.
The Growth Rate. The growth
rate is a function of the amount of control, rigor, and efficiency by which
failures are corrected. An arbitrary selection of a growth rate is not
recommended, since analytical techniques exist for predicting more realistic
rates as a function of equipment attributes and development program
characteristics. A range of acceptable growth rates may be established (e.g.,
from .45 desired, to .35 minimum), with the minimum rate used to trigger the
need for more aggressive action. (See next page.)
This figure illustrates two separate planning curves, the first with a starting MTBF equal to 10% of the predicted MTBF (i.e., a high-risk program) and the second with a starting MTBF equal to 30% of the predicted MTBF (i.e., a low risk program). Both curves use a slope of .49, and an initial test time of 100 hours in order to provide a sufficient period of operation for assuring that test procedures, facilities, and test operators are all functioning properly. The figure shows that approximately 6000 additional test hours are needed to achieve the required 300 hour MTBF, using the 10%, high-risk starting point.
Hardware To Be Tested. To determine the optimal number of equipment/systems to be allocated to reliability growth testing, trade-off analyses must be conducted. In order to optimize the use of equipment and test time, prereliability growth test analyses must be carefully performed in order to select high-risk equipment for emphasis in testing. If program constraints result in the need to test systems that are incomplete or that are not configured with realistic operational relationships among system elements, meaningful reliability measurements are not possible unless surrogate performance measurements at interfaces have been defined and implemented.
Test Time. The rate at which reliability growth can progress in calendar time is constrained by the rate at which weaknesses can be identified (affected by test time per unit and by the starting reliability) and by the speed of development and implementation of corrective actions (affected by engineering capability and management policy). Growth, therefore, is nonuniform and discontinuous.
The overall test hours needed for reliability growth testing can be estimated by using the reliability growth curve starting point, growth rate, and required MTBF. The calendar time needed to complete the testing is a function of the number of units under test and the overall test efficiency, influenced by:
- downtime for maintenance and repair of test chambers or debugging of new test chambers or setups,
- availability of spare and repair parts, and
- availability of resources for troubleshooting and