Automation of maintenance of construction and road machinery fleets using digital twins

Abstract

The fleet of construction and road machinery operated by Russian contractors retains a high proportion of units that have exhausted their normative service life, while maintenance planning still relies on schedules tied to accumulated engine hours. On-board condition monitoring and worksite data exchange are covered by current standards, yet no procedure has been established for assigning maintenance intervals from a remaining useful life forecast, and no quantitative comparison of maintenance strategies has been carried out at fleet level rather than for a single machine. The purpose of the study is to measure what the transition from preventive scheduled maintenance to maintenance driven by a fleet digital twin actually yields, and to decompose the resulting gain by source. An original simulation model of the annual operating cycle was built for a fleet of 48 earthmoving and road machines with a total utilisation of 68,270 engine hours; failure flows of five component groups were described by Weibull distributions, the annual cycle was reproduced by the Monte Carlo method over 10,000 replications, and factor contributions were separated by chain substitution. The technical availability factor rose from 0.913 to 0.945 while full unit costs fell from 1,116.6 to 781.2 roubles per engine hour. The share of sudden failures decreased from 100 to 30.0 per cent and average annual repair downtime from 135.3 to 82.2 hours per machine, which for the fleet is equivalent to releasing almost two additional units. Decomposition revealed a pronounced asymmetry: extension of service intervals contributed 0.0039 of the total 0.0322, while degradation recognition contributed 0.0293. Direct maintenance costs fell by only 4.0 per cent when moving from threshold condition monitoring to the digital twin, the remaining gain being formed entirely by reduced downtime losses. The value of a fleet digital twin is therefore created by redistributing maintenance events in time rather than by optimising the schedules themselves, and it is governed by the hourly cost of machine downtime together with the completeness of degradation recognition.