Airworthiness of aircraft. Part 2. Monte Carlo simulation of fleet performance history

Abstract

A computer program using the Monte Carlo technique to simulate aircraft performance in fleet operation (modelled in part 1 of this paper) is described. The technique follows the variation in the performance capability of each aircraft in the fleet over its service life. Arbitrary distributions or values can be specified for the input parameters of the stochastic model; in addition, the effect of certification and inspection procedures can also be studied. The output of the simulation includes the performance history of any specific aircraft, fleet performance distribution and statistics (in-service as well as just after overhauls) and the incident rates. Such a simulation of the single-engine climb gradient of a twin-engined turboprop aircraft leads to the following conclusions for a typical medium-haul airline fleet of 15 aircraft. The fleet mean rather than the standard deviation is generally sensitive to changes in the operating conditions. The time required for the fleet to relax from its new state to near in-service equilibrium conditions is found to increase from 4 months when maintenance is perfect to about 3 years when the maintenance efficiency (a measure of the extent of performance recovery in relation to new aircraft) is 25 %. It is found that the gradient considered acceptable at entry into service strongly influences the incident rates, but that the actual testing procedure adopted for clearing aircraft, such as single test, two-best-of-four etc., has hardly any effect. The incident rates are strongly affected by the maintenance efficiency and the flight scatter, moderately by the mean airframe and engine deterioration and the time interval between overhauls, and marginally by propeller deterioration. It is concluded that current airworthiness codes for engine-out take-off climb drawn up in the fifties are today generally conservative because of improved engine reliability, but may still be necessary for engines going through their 'learning' period.