Midnight at Sea: Saving a $1M/Day Operation with JMP and MCS (2026-US-PO-2909)

Executing a DOE 100 miles offshore on a moving vessel is rarely "textbook." When unverified epoxy coating threatened the integrity of a critical 4 million pound bolt connection, we faced a midnight decision: halt million-dollar-a-day operations or prove the system was safe.

Everything that could go wrong did. Initial testing suggested a special-cause wrench bias, only to discover a data-entry swap that reframed the crisis as high-scatter common-cause variation. As operational windows closed, our planned RSM DOE collapsed into a sparse comparative test. Working with non-ideal instruments amidst high winds and vessel heave, we implemented a "Master Bolt" for thermal drift compensation and an emergency MSA to validate our measurements.

When a wrench failure further corrupted our final run, we were left with imperfect, unbalanced data. Rather than relying on poor-fitting linear models, we leveraged JMP to build a digital twin reliability model. By integrating our field-derived preload distributions with high-fidelity load distributions from global riser analysis, we executed large-scale Monte Carlo simulations to evaluate the probability of flange separation. To ensure the integrity of the results, we analyzed the stability of the far-tail distributions using I and MR charts for convergence.

This session demonstrates how to maintain statistical rigor when "standard procedure" fails. We show how to transform sparse, messy field data into a probabilistic safety case, providing the engineering confidence required to continue critical subsea operations by coupling real-world "noise" with advanced structural simulations.