Reliability Modeling of Regional Cabled Observatories

Regional Cabled Observatories (RCOs) being planned today anticipate a twenty-five year operating life. Users expect the RCO to be operable a high percentage of the time, (i.e. to have high availability), while owners and operators require high reliability, in order to manage repair and maintenance costs. Meeting the requirements for power delivery, communications bandwidth, and control capability entails significant complexity in the RCO\´s underwater plant. The RCO is likely to combine Commercial Off-The-Shelf (COTS) equipment with equipment designed and qualified for use in commercial submarine telecommunications systems. The reliability of these components varies over a wide range; Mean Times Between Failure (MTBF) for individual functional blocks may range from 10⁸ to 10 hours. The repair and maintenance program for an RCO adds another level of complexity. Repairs may be possible only at certain times of the year and may take some time to initiate. Certain faults may be non-performance affecting and may be left for later repair. Assessing the reliability performance of such a system is beyond the scope of a traditional reliability analysis, which relies on various simplifying assumptions. To illustrate the impact of the system design and the maintenance plan on the expected availability of an RCO, computer models of two hypothetical system designs have been developed. Reliability block diagrams are used to ensure a consistent approach is applied to each design. The computer model includes logic which implements the reliability block diagram and nested loops which simulate a twenty-five year system life with random failures. A "Monte Carlo" approach is used to generate probability distributions of the expected availability and repair frequency. Variable input parameters include functional block failure rates, number of spares on hand, repair time, and the portion of the year in which repairs can be performed. The use of computer modeling is design neut- ral, considers the full system lifetime, allows a variety of repair strategies to be evaluated, and produces results that are meaningful for both end users and operators. Using the computer models, it is shown that an average availability of 97% is readily achievable with typical MTBF values and realistic repair assumptions. It is further shown that the selection of system design for the RCO has a limited impact on the overall reliability of the system, provided each design has similar failure rates. Finally, it is seen that the repair strategy has a significant role in determining overall availability.