One-sided predictive thresholds for aerospace electromechanical actuators: Monte Carlo and probability-plot methods

Electromechanical actuators (EMAs) on flight-critical surfaces require defensible, auditable thresholds for lane-level monitoring and fault isolation. The approach is broadly applicable and is demonstrated on an output-summed, multi-lane Sea Harrier EMA with continuous lane monitoring and voting. Monte Carlo sampling of motor and sensor tolerances drives simulations to estimate peak lane disparity distributions for current, potentiometer, and tachometer signals across Mach 0–1 and the commanded-deflection envelope under inertial and aerodynamic loading. Normal and lognormal probability plots provide visual diagnostics and guide tail-model selection between two and three parameter lognormal forms. The workflow aligns the classical straight-line probability-paper practice with automated lognormal fitting to set one-sided predictive thresholds at a 10⁻⁴ nuisance-disconnect (healthy false-alarm) target per operating point, while quantifying model-choice effects (Linear, 2-parameter lognormal, 3-parameter lognormal) and their impact on conservatism at very low probabilities. The procedure produces traceable, tool-based thresholds that suit industrial needs for repeatability and auditability in voter-logic architectures. Results show that three-parameter lognormal modelling can add useful conservatism in the extreme tail. Automated lognormal fitting also provides a practical foundation for AI-based, predictive threshold setting that adapts to evolving operating conditions. Overall, the workflow yields certifiable thresholds for aerospace applications, where transparency of assumptions and protection against rare events are as critical as detection performance.