Theory and Practical Experience of a Digital Twin for Optimal Operation and Maintenance of Belt Conveyors

Alarge part of the bulk material transport in mining is carried out by belt conveyors. This will increase in the future. The correct design, operation and maintenance are of decisive importance for their economic operation. Recent developments such as the use of a digital twin of the conveyor can make a significant contribution here. This article is intended to present the theory and operational practice of the digital twin “BeltGenius” for conveyor systems developed by J. M. Voith SE & Co. KG, Crailsheim/Germany.

---

---

1  Introduction

---

A large part of the bulk material transport in mining is carried out by belt conveyors. This will increase in the future. The correct design, operation and maintenance are of decisive importance for their economic operation. Recent developments such as the use of a digital twin of the conveyor can make a significant contribution here. Since the term “digital twin” has almost become an inflationary buzzword that everyone understands differently, this article is intended to present the theory and operational practice of the digital twin “BeltGenius” for conveyor systems developed by J. M. Voith SE & Co. KG, Crailsheim/Germany.

---

2  Basics of the calculation model

---

2.1  Analytical calculation of the main resistance

---

The core of the digital twin is a mathematical simulation of the belt conveyor, comparable to the relevant calculation programs with which belt conveyors are usually designed. These calculation programs are based on formulas that supply the local movement resistances along the course of the belt depending on the respective operating parameters (1, 2).

The digital twin differs from conventional calculation programs as follows:

1. While the belt conveyor calculation programs are used “offline” for various operating states and loads as required, the digital twin is operated with real operating data over longer periods of time – either offline with historical operating data or online parallel to ongoing operation.
2. While the belt conveyor calculation programs work with empirical values for previously unknown values of the parameters in the calculation formulas, with the digital twin the parameters are automatically determined from the comparison between calculated and measured drive power using optimization processes.
3. Due to the high quality of the simulation, the digital twin allows quantitative statements on the load on all components and the evaluation of the components with regard to the load collectives achieved and their share of the energy consumption.
4. The digital twin provides information on the optimal operation of the system, evaluations of the energy efficiency achieved in operational practice and detects deviations resulting from physical changes to the system.

The digital twins of different providers differ in the complexity and accuracy of the mathematical replica. J. M. Voith SE & Co. KG, Crailsheim/Germany, sees the advantage of its model in the analytical simulation of the main resistance in contrast to the virtual friction coefficient that is otherwise commonly used (1). Its analytical model is based on many operational measurements and includes the influence of the local belt tensile force on the material flexing, the specific properties of the belts on the rubber flexing, and the ambient temperature on the heating of the belts and the associated change in the movement resistance.

---

2.2  Allocation of the load

---

Since the lifting work has a significant influence on the power requirement – in terms of energy, 1 m lift corresponds to about 50 m horizontal transport – and a belt conveyor has always inclined sections (Figure 1), the movement resistances are very different in the individual sections of the belt conveyor.

Figure 2 shows two different load distributions with the same average load on a belt conveyor with four sections. It goes without saying that in the lower case – despite the same average load – the power requirement is considerably higher than in the upper.

Accordingly, the position of the belt scale must be known for the digital twin and the loading of the upper run must be correctly synchronized with the associated motor power at every point in time.

---

2.3  Consideration of different belts

---

The belts have the greatest influence on the moving resistance of a belt conveyor due to the internal damping against the deformation of the rubber (indentation rolling resistance and belt bending resistance) (Figure 3).

If belts from different manufacturers and production batches are used in the same belt conveyor, they can differ significantly in terms of their load-dependent moving resistance (Figure 4).

In these cases – similar to the loading – the accuracy of the simulation depends on the position of the individual belts with their respective loading being known at all times. For this purpose, an RFID-based detection system was developed for BeltGenius, which records the start of each belt when it passes an antenna and synchronizes it with the load.

---

2.4  Setting up and initializing the digital twin

---

At the beginning, a structural data set of the belt conveyor is created, which contains the height of the belt along the conveyor, the position of the pulleys, drives, loading and unloading point, belt scale, tensioning device and the sensors for speed, belt tension measurement and belt detection as well as technical data of the main components. In principle all the Information that is also required for a classical belt conveyor calculation.

The initialization, i. e. the determination of the parameters for the exact description of the operating behavior, takes place on the basis of several time periods, which as far as possible include all load states and ambient temperatures occurring during operation. The start parameter set is then varied by a multi-stage optimization process until a specified iteration limit is achieved. The target function for this optimization run is the least squares sum QS from the difference between the calculated and measured motor power over all measuring points according to the following formula:

---

3  Examples of evaluations

---

3.1  Achieved precision

---

Figure 5 shows the reproduction quality for a belt conveyor in the Garzweiler open pit mine (nominal capacity 37,500 t/h,
v = 7.5 m/s, B 2700 St 4500 16:8, motor power 2 x 2,000 kW, 850 m center distance).

---

3.2  Assessment of the energy efficiency of the belts used

---

Most belt conveyors are equipped with the belts from a single manufacturer, which is why the individual belt sections differ significantly less in terms of energy efficiency than shown in figure 4. The high level of replication accuracy in figure 5 required the specific properties of the individual belts to be taken into account. This is possible with BeltGenius if a belt detection system is used. In these cases, the respective parameters are also determined for each belt by an automatic optimization process. Figure 6 shows the determined energy efficiency of the ten belt sections from the example above. It should be noted that compared to the results from 2009 shown in figure 4, significant improvements in the energy consumption of the belts can already be seen.

With the digital twin, very precise forecasts can be made as to how the belt conveyor would behave under different operating conditions. It is possible, e. g., to calculate what the energy consumption would be if other belts were used. For the system considered here, the time range shown in figure 5 with regard to energy consumption was determined for the following three variants:

1. current status: existing mix of ten different belts;
2. best-case scenario: exclusive use of the best of the ten belts;
3. worst-case scenario: exclusive use of the worst of the ten belts.

The results are are as follows (Figure 7):

• With the actually installed combination of ten belts, the energy consumption is 6,733 kWh or 115 Wh/(t*km).
• If only the best of these belts were installed, the energy consumption would be 6,195 kWh or 105 Wh/(t*km).
• The worst belt, on the other hand, would drive the power consumption to 8,328 kWh or 142 Wh/(t*km) and pushes the installed capacity to its limits.

It should be noted that these results include 1,683 kWh of energy consumption, which cnnot be influenced by the belts.

---

3.3  Evaluations regarding system utilization and energy consumption

---

BeltGenius offers a number of very useful overviews showing utilization and specific energy consumption of different time periods can be seen. One time range is the operating time of the conveyor system from start-up to shutdown. Figures 8 to 11 show this as an example for a Chilean copper mine over the period of one year.

---

3.4  Specific calculation results for individual time periods

---

With the currently valid set of structure and status parameters, a number of parameters are calculated for each time period and stored in topic-specific files, from which the desired visualizations are generated. Figure 12 shows some examples of this.

---

3.5  Lifetime estimation of components

---

For most components, there is currently no operational load measurement available or not used for economic reasons. The only evaluation criterion is often the installation time. However, BeltGenius can seamlessly record the relevant forces or moments as well as paths and angles of rotation for all moving parts. If the relationship between stress and service life consumption is known, service life forecasts can be automatically generated from this. In most cases, however, this relationship is not known in advance or only under operating conditions that are significantly different from the current ones. Since such predictions are of great economic importance, it makes sense to look for the respective connections for the various components. In addition to the date of installation and removal of the components, a record of all possible influencing factors is required in order to find a correlation between different calculation approaches and the real operating times achieved.

The evaluation of the roller bearings of the six pulleys according to (5) is presented here as a simple example of a theoretical service life estimation with BeltGenius. Figure 13 shows the course of the axle loads of these six pulleys for a real-time range of 10,000 s.

The calculated service life according to (5) for the respective bearing load (= half axle load) can be calculated for each individual second. The summation of the reciprocal values then results in the lifetime consumption for this time range.

Pulley 2 and 3 have the highest axle loads, but are also fitted with different roller bearings than the other pulleys. Figure 14 shows the percentage of lifetime consumption for those 10,000 s.

The bearings of pulley 4 and 5 are practically fatigue-resistant (curves for pulley 4 and 5 are superimposed on the x-axis).

Other possible influencing variables that were not taken into account here but are available as data: Ambient temperature, number of start-ups and standstill times and thus the cooling times. This theoretical calculation could then be improved with the operating times actually achieved. Other components can be viewed in a similar manner. In addition to the belt speed, the radial loads from the belt and loading weight as well as the orthogonal components of the local belt tension are available for the idlers, for the belt connections the changing stress caused by the belt tension, etc.

---

4  Practical examples of detected deviations and errors

---

4.1  Coal mine in Australia (underground)

---

For a coal mine in Australia, a period of ten weeks was analyzed using BeltGenius. The reason for this was persistent problems with intermittent overloads, which led to shutdowns of the fully loaded conveyor. Each of these overload shutdowns led to a longer downtime during which the conveyor had to be partially emptied manually before it could start up again.

The analysis using BeltGenius revealed errors in the control of the drives and the belt tensioning device. The cause was that the belt tension measuring devices behind the two tripper drives did not supply any measured values when 200 kN was exceeded, which the controller interpreted as too low belt tension, reduced the power on these drives and further increased the belt tension. In addition to the unnecessary overload shutdowns caused by this, a pulley shaft also broke.

Figure 15 shows the comparison of the belt tensile force calculated by a consultant for an even loading of 3,700 t/h in comparison with the calculation by BeltGenius for a real average loading of 2,640 t/h.

You can see that with the real (lower) load, the maximum belt tensile force of approximately 360 kN was around 2.6 times the 137 kN calculated by the consultant.

Since BeltGenius can very precisely determine the course of the belt force based on the belt force measured on the tensioning winch, it is possible to dispense with the unreliable belt tension measuring devices behind the two tripper drives.

Figure 16 shows the sometimes very poor power distribution between the five motors (each with a rated power of 440 kW). The control unnecessarily reduces the power of motors 4 and 5, particularly when there is a higher power requirement, which causes the overload shutdowns.

---

4.2  Copper mine in Chile (open pit)

---

A whole year was evaluated for this copper mine. Of the three most important deviations – temperature-dependent offset error of the belt scales in the range of 5 to 8 % of the nominal power, poor power distribution, faulty start-up processes – only the poor power distribution of the three slip-ring motors (3 x 1,250 kW) is explained here (Figure 17).

Especially with a low total power < 600 kW, the contribution of motor 3 is particularly small and is taken over by motor 2. This is disadvantageous for the transmission of drive 3, since too little load leads to damage to the roller bearings. BeltGenius can be used to determine the gradient of the motor characteristics and the effective diameter of the drive pulleys and to simulate measures to improve the power distribution, e. g., specific information on changing the continuous slip resistance.

Incidentally, the analysis of the system with regard to belt safety, motor power and filling cross section showed that an increase in output from the current 7,000 to 9,000 t/h is possible without any problems (Figure 9).

---

4.3  Copper mine in Chile (underground)

---

Four weeks of operating time were analyzed here and the following deviations were found:

• Belt scale signal has a time offset that jumps between 0 and 120 s.
• Belt scale signals are temporarily too low by up to 800 t/h.
• The drive control of the direct drives produces vibrations, which explain the premature failure of a drive pulley after only one year of operation.

In particular, the failure of a drive pulley after only one year of operation poses a major problem for the customer, since the cause has not yet been determined and there was uncertainty about the future performance of the system.

BeltGenius gave indications of a very volatile power distribution and oscillations in the drive power. Measurements then initiated found the cause in the unfavorable design of the drive control. Figure 18 shows a section of these measurements for one of the four drive motors (start-up process, empty conveyor system).

---

5  Summary

---

The use of a digital twin for belt conveyors, which can be adapted to any belt conveyor with little effort and reproduces them exactly using real measured values – enables a quick overview of the condition, efficiency and utilization of the system. In this way, starting points for improvement can be identified and quantified. This is a significant advancement compared to the usual calculation and simulation tools.

In the fight against climate change, efforts to reduce energy consumption must also be stepped up in mining. Conveyor belt systems are much cheaper in terms of energy consumption compared to the use of trucks and are more economical overall from a transport distance of approximately 3 km. There is still considerable potential for improvement in the design of the belt conveyors and the selection of the components. This can be shown and realized by BeltGenius.

BeltGenius offers an even greater savings potential as an instrument for the optimal operation of the systems. In the example of Chilean opencast copper ore mining discussed here, the average savings potential by avoiding production gaps and correspondingly lower – but more even – production capacity is around 20 %. Significantly greater savings would result if a constantly high conveying capacity could be achieved and the operating time would be reduced by the avoided production gaps. That sounds trivial, because every miner always strives for a high output anyway. However, it is much more economical to strive for a constantly high level of utilization instead of trying to make up for gaps in funding and downtimes with temporary peak performance. Here, BeltGenius can show the management the economic effects when making decisions about improving the operational organization.

Another perspective for the use of a digital twin results from the fact that the expertise for optimal operation and maintenance of conveyor systems can be assumed by fewer and fewer operators. A monitoring system for belt conveyors is needed that increasingly replaces this expert knowledge. However, learning, maintaining and expanding such a system requires close cooperation with the operator.

---

---

---