FAA source text, reformatted for reading. Page boundaries and paragraph numbers are retained.
7.8.2 Single failures in combination with operational or environmental conditions leading to catastrophic failure conditions are in general not acceptable. However, single failures do not need to be assumed in combination with operational events or environmental conditions that are extremely remote or that occur outside the normal flight envelope defined in AC 25.671-1. Other cases that are properly justified may be accepted on a case-by-case basis by the certifying authority. In limited cases where a non-redundant system provides protection against an operational or environmental condition (for example, a fire protection system in the cargo compartment comprised of detection and suppression functions) any single failure that results in the loss of the protection function should meet the criteria associated with the major failure condition classification, to ensure adequate system reliability and development assurance.
7.9 Justification of Assumptions, Data Sources, and Analytical Techniques.
7.9.1 Any analysis is only as accurate as the assumptions, data, and analytical techniques it uses. Therefore, to show compliance with the requirements, the underlying assumptions, data, and analytic techniques should be identified and justified to assure that the conclusions of the analysis are valid. Variability may be inherent in elements such as failure modes, failure effects, failure rates, failure probability distribution functions, failure exposure times, failure detection methods, fault independence, limitation of analytical methods, processes, and assumptions. The justification of the assumptions made with respect to the above items should be an integral part of the analysis and summarized in the safety analysis. Assumptions can be validated by using experience with identical or similar systems or components with due allowance made for differences of design, duty cycle, and environment. Where it is not possible to validate a safety analysis in which data or assumptions are critical to the acceptability of the failure condition, extra conservatism should be built into either the analysis or the design. Alternatively, any uncertainty in the data and assumptions should be evaluated to the degree necessary to show that the analysis conclusions are insensitive to that uncertainty.
7.9.2 Where adequate validation data is not available (for example, new or novel systems) and extra conservatism is built into the analysis, then the normal post-certification in-service follow-up may be performed to obtain the data necessary to alleviate any consequence of the extra conservatism. This data may be used, for example, to extend system check intervals.
