Why the Torpedo Engagement Kill Chain Underestimates the Probability of Kill and How to Fix It

摘要

For the past several years, the U.S. Navy has assessed the performance of lightweight torpedoes (LWT) as a critical issue. The current de facto assessment is based primarily on the testing of Recoverable Exercise Torpedoes (REXTORP) and Exercise Torpedoes (EXTORP) LWTs. Because a kill never actually occurs and the REXTORPs are only partial runs, the U.S. Navy assesses LWTs by evaluating the following three events of the engagement kill chain: System Reliability (R_(sys)), Torpedo Reliability, and Torpedo Effectiveness. These events are assessed independently, and they are averaged to obtain probability values for the kill chain factors. These probabilities are then multiplied, and the result is reported as the Probability of Kill (P_k). This assessment is problematic because the kill does not actually occur. The grading of R_(sys) is linked to proficiency standards rather than to mission success. Dependencies across events in a torpedo run are not accounted for, such as an unsatisfactory torpedo grade that would otherwise allow for continuation of the kill chain and a potential mission success (e.g., target hit). Averages combine data from a wide range of tests with conditions that can significantly affect performance of the events. This paper proposes a multi-state engagement kill chain to account for such degraded events, including compensatory effects associated with functional redundancy. This would provide a holistic approach with a focus on mission success similar to the assessment of Heavyweight Torpedoes (HWT). A hypothetical example that uses this more realistic assessment indicates that the LWT R_(sys)and P_k are significantly higher than currently reported.