Optimisation of maintenance strategies for belt conveyors in Rheinbraun's
Report 2003.TL.6800, Transport Engineering and Logistics.
RWE Rheinbraun AG operates three opencast mining plants Garzweiler,
Hambach and Inden. These mines are located in the area between the cities
Köln, Düsseldorf and Aachen.
The lignite and overburden are transported mainly in and between the mines
with belt conveyor systems. The lignite is further transported from the
mines to the power stations with trains (Garzweiler, Hambach) or belt
conveyors (Inden). The annual amount of conveyed material is approximately
100 million tons lignite and 500 million tons overburden. There are about
250 km belt conveyors (axis-to-axis distance) operational. The transport
performance is 10 billion ton kilometres per year.
Bucket wheel excavators mine the lignite and overburden. The capacity of
the conveyors is adapted to the capacity of the excavators.
The conveyor belts represent a large amount of invested capital. Given a
total length of approximately 500 km belts installed in the conveyors and
an average lifetime of seven years, the annual costs for replacement are
about 35 million euros.
The belts are used under severe circumstances. Rocks up to one cubic meter
fall from heights of 7 till 12 meters on the belts. These impacts cause
damages to the belts and shorten their life times. Furthermore belt
damages cause downtimes of the conveying system by means of failures and
These damages have to be repaired periodically to prevent further growth of
the damages, failure and downtime of the production process. Downtime costs
are estimated at 150,000 euro per working shift (8 hours).
The total costs of maintaining the conveyor belts include manpower, use of
equipment, materials and production losses. These total costs are a
multiple of the 35 million euros mentioned earlier for the annual costs
for replacement. The costs for belts compared to the total costs of a
conveyor are in the range from 40% till 70%.
Savings in the order of tens of millions of euros can possibly be saved
annually by optimizing and controlling all maintenance.
Objective of the research project is to find an optimal maintenance policy
with respect to economics of the mining plant (maximum profit).
Optimization needs compromising the following objectives:
- minimum maintenance costs (repair- and replacement costs),
- minimum downtimes and
- maximum system's availability.
Primary target of the assignment is to develop maintenance strategies,
which are focused on minimization of the costs of the plant. Secondary
target is to present additional information like belt lifetimes, average
costs per belt, etc.
There are several maintenance policies known from literature. These can be
classified into three groups: age-based policies, condition-based policies
and reliability-based policies. Reliability-based policies are the most
sophisticated maintenance policies. Rheinbraun currently uses
condition-based policies, but not the same for the three mines. The
differences are that the policy of Inden is focused on maximum
reliability, that of Garzweiler on maximizing belt life and Hambach
focuses on having the repairs carried out in the workshop. These differences
are based on local preferences, and are not necessarily the most economical
policies. Evaluation of the most economical maintenance policy for a
complicated belt conveyor system requires a computer model such a
discrete-event simulation model.
A discrete event simulation model for evaluation of various strategies for
maintenance of belt conveyors has been successfully developed. The model
has been coded in Borland Delphi, an object oriented programming language.
The program has a graphical user interface and a Windows-look. The program
uses the simulation control and event handling mechanisms of TOMAS, a
simulation tool developed by and available from the Delft University of
For demonstration purposes and testing of the capabilities of the model a
number of simulation runs (experiments) have been carried out.
A base (reference) experiment and eleven variants are carried out to study
the effects of:
- the thresholds for condition based rules,
- the damage growth speed,
- the frequency of inspection intervals,
- the minimum replacement length and
- the number of available vulcanisation workers.
Using the currently available best estimates for the input parameters the
following results have been obtained:
The experiments are carried out using estimated distribution functions for
damage occurrence frequency, size, position and growth. The quality of the
input and output can be improved by using distribution functions
constructed with measured data from practice. It is recommended to collect
this measured data in a centralized database system.
- The maintenance variants studied show differences of over 100 million
- Downtime costs are the most dominant cost factor (approximately 85%).
- Applying a preventive maintenance strategy raises the costs with 15
percent. The corrective maintenance variant lowers the costs with 17
- Conditions for an inspection maintenance model are damage growth, the
ability to compare various options for decision-making in terms of
reliability, production loss and maintenance costs.
- The effect of the length of new belts on the total costs is not
- The effect of the number of available vulcanisation workers on the
total costs is not significant (±1%).
Besides the input data the model itself can be improved with the following
Further optimization can be achieved by developing a top-down optimization
model for the entire mining plant taking aboard simulation models for
separate equipment such as for the belt conveyors.
- Extend the current model to the entire mining plant.
- Incorporate a decision-making support procedure.
- Model the effect of the type of conveyed material with respect to
damage occurrences and growth.
- Replace the current random function for repair costs with a function
that takes the real expected repair costs into account. This will be a
function of the damage properties.
- Model a failure mechanism for cover wear.
- Replace the current simple model for damage growth with statistical
distribution functions constructed using data from the central database
system or if available with theoretical models.
- To investigate the effect of replacing short (virtual) belt lengths
use an algorithm to calculate the most cost-effective start and end
- To investigate the effect of the pool size of vulcanisation workers
take into account the costs for unassigned vulcanization workers.
Reports on Transport Engineering and Logistics (in Dutch)
, TU Delft