2006:009 CIV
MA S T ER’S TH E SI S
On Predicting Fuel Consumption and Productivity of Wheel Loaders
MATS BOHMAN
MASTER OF SCIENCE PROGRAMME Engineering Physics Luleå University of Technology Department of Mathematics
2006:009 CIV • ISSN: 1402 - 1617 • ISRN: LTU - EX - - 06/9 - - SE
Abstract Low fuel consumption is becoming more and more important as a selling argument as fuel prices keeps rising. To be able to better show their advantage in this area, Volvo Wheel Loader need to develop the selling tool SiteSim in terms of fuel consumption predictions. This means modelling the actual work done by the wheel loaders. A good physical model would produce the results wanted. This modelling is however very difficult due to the interaction between the hydraulic system and the drive train. To overcome this problem, a method of splitting up the work cycles of a wheel loader into phases has been developed. The phases are defined by a specific type of work, for example filling bucket or reversing from bank. This reduces the effecting parameters and makes physical modelling possible in some phases. It also makes statistical modelling easier and more exact. The statistical models yields good results if the user is experienced in the work the simulated wheel loader are to perform. The physical models produce results with 10-20% lower fuel consumption than tested value. Better models of energy losses in drive train would probably correct some of these errors.
Sammanfattning Allt eftersom br¨ anslepriserna g˚ ar upp blir en l˚ ag br¨anslef¨orbrukning ett allt viktigare f¨ors¨ aljningsargument. F¨ or att b¨attre kunna visa sitt f¨orspr˚ ang p˚ a detta omr˚ ade beh¨ over Volvo Wheel Loader f¨ orb¨attra sitt f¨ors¨aljningsverktyg SiteSim vad g¨aller br¨anslef¨ orbrukningsber¨ akningar. En bra fysikalisk modell skulle ge de o¨nskade resultaten. P˚ a grund av kraftdelningen mellan hydraulik och drivlina a r denna modell sv˚ ar att realisera. En metod som, ¨ genom att dela upp arbetscyklerna i mindre bitar, undviker detta problem har utvecklats. Dessa mindre delar kallas faser och defineras av en specific typ av arbete, till exemplel fylla skopa eller backa fr˚ an banken. Inom varje fas kommer p˚ a detta s¨att f¨ arre variabler p˚ averka systemet och det blir i vissa faser m¨ojligt att model¨ lera fysikaliskt. Ovriga faser blir genom uppdelningen ocks˚ a l¨attare att modellera statistiskt. De statistiska modellerna ger goda resultat om anv¨andaren ¨ar van arbetet som ska utf¨ oras. Resultaten fr˚ an de fysikaliska modellerna visar p˚ a en 10-20% l¨agre br¨anslef¨ orbrukning ¨ an resultat fr˚ an testk¨orningar. B¨attre modellering av f¨orlusterna i drivlinan kommer troligtvis minska dessa fel n˚ agot.
Acknowledgements I would like to express my sincerest gratitude towards all the people that have supported me though this thesis. I would like to express special thanks to: • My supervisor at Volvo Wheel Loader, Stefan Pettersson, for his guidance and support, it has been invaluable. • Tomas Gunnarsson, my examiner at LTU. • Sven-˚ Ake Carlsson, for always finding time to answer questions and provide guidance in areas concerning power distribution. • Reno Filla, for his support in building the model. • Esko Bjurman and Pauli Hanssen for acting as operators during the testing weeks. • Stefan Asplund, for installation and help with testing equipment. • The whole department of UT, for their willingness to answer all possible questions, interest in my work and for making me feel welcome at Volvo. Finally, since this thesis ends my time at LTU, I would like to thank all the teachers and fellow students that has made my time in Lule˚ a such a great time. G¨oteborg, October 2005. Mats Bohman
Contents 1 Background 1.1 Volvo Hauler Loader Business Line HLBL . . . . . . . 1.2 SiteSim . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Reason for upgrading the SiteSim wheel loader model 1.4 Problem description . . . . . . . . . . . . . . . . . . . 1.5 Scope of master thesis . . . . . . . . . . . . . . . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
1 1 1 2 2 2
2 Basic wheel loader theory 2.1 Background . . . . . . . . . . . . . . . . 2.2 Main uses of large machines . . . . . . . 2.2.1 Short cycle . . . . . . . . . . . . 2.2.2 Load and carry . . . . . . . . . . 2.2.3 Timber handling . . . . . . . . . 2.3 Power distribution . . . . . . . . . . . . 2.3.1 Engine . . . . . . . . . . . . . . . 2.3.2 Hydraulic system . . . . . . . . . 2.3.3 BSS - Boom Suspension System 2.3.4 Drive train . . . . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
4 4 4 4 5 5 5 5 8 8 9
3 Hypothesis 3.1 The difficulties . . . . . . . . . . . . . 3.1.1 The operator . . . . . . . . . . 3.1.2 The dynamics of a wheel loader 3.2 Purposed solution . . . . . . . . . . . 3.2.1 Phases and parameters . . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
11 11 11 11 12 12
4 Gathering data 4.1 Pretest thoughts . . . . . . . 4.2 Test vehicle . . . . . . . . . . 4.3 Sensor set up . . . . . . . . . 4.4 Performed tests . . . . . . . . 4.5 Obtaining data from test logs
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
16 16 16 17 18 18
5 Resulting model 5.1 Filling bucket phase modelling . . . . . . . . . . . . . . . . . . 5.2 Reverse from bank and reverse from receiver phases modelling . 5.3 Emptying phase modelling . . . . . . . . . . . . . . . . . . . . . 5.4 Transportation phases modelling . . . . . . . . . . . . . . . . . 5.4.1 Statistical modelling . . . . . . . . . . . . . . . . . . . . 5.4.2 Physical modelling . . . . . . . . . . . . . . . . . . . . . 5.5 Summing up the phases . . . . . . . . . . . . . . . . . . . . . . 5.6 Parameters in effect . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
21 21 22 22 22 23 23 24 24
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
6 Conclusions and future work 6.1 Model accuracy . . . . . . . . . . . 6.2 Requirements for good results . . . 6.3 Effects of boom suspension system 6.4 CAN-fuel signal verification . . . . 6.5 Future work . . . . . . . . . . . . . A Statistical background
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
26 26 26 27 27 27 30
Chapter 1
Background 1.1
Volvo Hauler Loader Business Line HLBL
This thesis was ordered by Volvo Wheel Loader Business Line (WLBL) in Eskilstuna. During the time the work was done it was decided that WLBL and Articulated Hauler Business Line (AHBL) were to form the new Hauler Loader Business Line (HLBL). The thesis was done at UT (Sales engineering & Marketing support) within WLBL. This department includes Sales engineers, Product specialists, Market communication and Product administration & System development. The Sales engineers support dealers with technical expertise in sales and marketing situations, e.g. customer visits, sales training etc. The Product specialists support both dealers and WLBL with competence in machinery, applications and attachments. Work involves both product and sales development. Market communication produce marketing material, develop sales tools and manages events. Product administration & System development coordinates basic product information, administrates options and attachments and develop sales systems.
1.2
SiteSim
Both as a retailer and an entrepreneur it is good to be able to estimate how long time and how many machines are needed to do a specific job for a potential customer. Traditionally, this was done using a performance manual for each of the machine types used. Using the performance manual and a map of the potential site you were able to estimate these figures. The solution was crude and required hard work. Though the performance manuals are still available a new tool has also been developed. The new tool is SiteSim. SiteSim is a computer program which can do all what an experienced user of the performance manuals can do, and more. The user is able to build up the site, complete with dig phases and haul segments. It is also possible to set parameters such as fuel cost, interest cost, operator cost, work schedules, etc. Using these parameters the program provides you with the optimal number of machines, the
1.3. Reason for upgrading the SiteSim wheel loader model
2
cost of production and the required time to reach a production target. The user may also investigate the effects of using different equipments on the machines.
1.3
Reason for upgrading the SiteSim wheel loader model
Fuel consumption has been modelled with data collected over long times from customer sites using the D-series wheel loaders. When the E-series was introduced the SiteSim-model was simply corrected with a constant with accordance to the performance of the E-series opposed to the earlier D-series. This yields true results in most cases but might give a little better or worse performance depending on the task. Another problem is that the resolution of terrain quality is not correct. You have the possibility to choose between good, bad or average. However, bad terrain can be muddy, with wheels sinking several decimetres into the mud, making it very heavy to move the machine, resulting in low speed. It can also be very rocky, with stones sticking up several decimetres, forcing you to drive very slowly. Both cases effects production in the same way, slow speed decreases the number of cycles per hour, which in turn decreases production. If we examine fuel consumption, the difference comes into notice. In the muddy case, we force the engine to work heavily to propel the loader, leading to a high fuel consumption. In the case with large rocks, the terrain forces us to drive slowly without forcing the engine to work hard which leads to a lower fuel consumption. This difference is not modelled. Also since the calculations are based on data from customer machines, any new machines performance is only available when that specific model has been on the market for quite some time. This means that you can not use SiteSim as a good selling tool for new models.
1.4
Problem description
A new model of calculation is to be created. It shall be based on the current wheel loader series, it needs to be easily upgraded when a new wheel loader series reaches the market. The new model should also have a finer resolution of parameters in ground conditions in terms of ground structure and rolling resistance.
1.5
Scope of master thesis
This master thesis is limited in time to 20 weeks which rises the need for limitations in project scope. When planning the project the following goals was set up: • Core model • Easy add functionality or upgrade to new model series • Implementations in MATLAB
3
Chapter 1. Background
• One application • Handle flat ground as well as slopes • Examine the effect of Boom Suspension System (BSS) on fuel consumption • Examine the effect of BSS on productivity?
Chapter 2
Basic wheel loader theory 2.1
Background
The wheel loader is a further development of the standard tractor. In 1954 BolinderMunktell AB realized that, by placing the bucket in the back over the large wheel axle, a higher load could be lifted. The back-end loader was born (figure 2.1). Bolinder-Munktell became BM Volvo, Volvo BM, VME (Volvo-Michigan-Euclid) and now Volvo CE. The back-end loader has evolved into the wheel loader with all-wheel drive and frame steering for manoeuvrability. The sizes range from 8 to 52 metric tonnes which enables multiple uses from forklift loading and snow clearing to rock and timber handling.
2.2
Main uses of large machines
The large machines generally include all machines larger than 23 metric tonnes (the L150E, L180E, HL180E, L220E and the L330E). These are, depending on model and equipment, mainly used for moving heavy loads of sand, gravel, rock or timber. In most cases the work follows a repeating cycle, starting at the dig phase where it fills the bucket etc., it then transports the material to the receiver (hauler, truck, crusher etc.), empties the bucket and returns to the dig phase.
2.2.1
Short cycle
A short cycle loading is when the receiver is placed adjacent to the dig site. The transporting distance should only be long enough to give the boom enough time to lift the bucket over the side of the receiver (see figure 2.3). A longer transporting distance decreases productivity since a smaller fraction of working time is put into digging.
5
Chapter 2. Basic wheel loader theory
Figure 2.1: The 1954 Bolinder-Munktell H10 back-end loader loading a hauler
2.2.2
Load and carry
In a load and carry cycle the receiver is not adjacent to the dig phase resulting in a longer transportation with the material in the bucket. The receiver can be a hauler, truck or pocket (crusher, hopper etc.). In the case of a pocket receiver, the unloading height will usually be quite low.
2.2.3
Timber handling
Timber handling does not usually show the same repeating pattern as material moving tasks. It involves a lot of transportations and quite a lot of work fixing the piles and sorting the timber.
2.3
Power distribution
The Volvo wheel loaders are diesel engine driven. The way the power is distributed between hydraulics and drive train depends on how much power is requested from the different systems.
2.3.1
Engine
The engine in Volvo’s large wheel loaders is a turbocharged, electronically controlled diesel engine. The electronic control handles fuel injection and tries to deliver the
2.3. Power distribution
Figure 2.2: A L180E in timber handling
6
7
Chapter 2. Basic wheel loader theory
Figure 2.3: Loading a hauler in a short cycle
2.3. Power distribution
8
torque requested by the machine. The requested torque depends on several factors, one is keeping the engine speed at a, by the operator, requested level. This is different from a normal car engine and means that by pressing down the accelerator pedal you request a higher engine speed, not necessarily a higher power output as is the case in a car engine. On the other hand if a higher load is put on the engine this does not necessarily effect the engine speed. The reason why the engine is speed controlled is that since the load on the engine can differ largely in a short time, due to the hydraulics out-take, a power controlled engine would be very hard to operate and would probably come to a halt very often. As an operator you do not feel this difference when moving. The reason is that you react the same way. If you want to accelerate you press the accelerator pedal. This allows the engine to reach a higher engine speed. A higher engine speed usually results in an acceleration (not always due to the torque converter 2.3.4). Diesel engines have a governed speed, an engine speed where you limit the speed to go up too high by reducing the fuel injection. If this wasn’t the case the engines could keep going up in engine speed and risk breakdown. When the accelerator pedal is fully pressed this built in speed limit hinders the engine to over speed. By only pressing down the accelerator pedal some bit you change this limit rpm to a lower value. There is also a run-out speed, a higher limit, at which injection is lowered so that no usable torque is delivered. The efficiency of today’s diesel engines is very high, with inter cooler and turbo it is about 43%. It could actually be a few percentage units higher if we could ignore today’s environmental regulations. The losses are exhaust (30%), cooling (20%) and friction (the rest). It is the exhaust part that could be lessened if no regulations had to be followed but this would result in higher levels of nitrogen oxides (N Ox ) in the exhaust.
2.3.2
Hydraulic system
The hydraulic system delivers power to the working hydraulics (lift, tilt and optional functions if available), steering, servo, brake and to the hydraulic motor of the cooling fan. The hydraulic pumps are axial piston pumps with variable flow. This means that when neither flow nor pressure is needed the flow is reduced to a minimum and almost all power can be used by the drive train. However, the system is always kept at a stand-by pressure that allows quick responses to operator commands. This system provides high performance by minimizing power losses.
2.3.3
BSS - Boom Suspension System
The boom suspension system is an optional function on Volvo’s wheel loaders that allows the hydraulic system operating the lifting cylinder to work as a suspension system for the boom. This means that transportations over uneven ground can be done in a more comfortable way.
9
Chapter 2. Basic wheel loader theory
Figure 2.4: L180E with part names
2.3.4
Drive train
The power from the engine is transferred and geared down in several steps. Closest to the engine is the torque converter followed by (in this order) gearbox, drop box, differential and finally hub reductions. Also affecting the driving force is the size of tires.
The torque converter The torque converter is easily described as two propellers in a closed housing, filled with oil. The engine is connected to one propeller and when turning this propeller sets the water in motion. The other propeller picks up the water motion and delivers this power to the rest of the drive train. Obviously this system can only deliver a smaller or at best an equal torque as the one put into the system. However by designing the pool and the propellers in a special way we can make this construction deliver a higher torque (at a lower rotational speed) than has been put into the system, that is, make it work as a gear down. The torque converter, as opposed to a mechanical gear down, allows different ratios between speed out and speed in (nout /nin ). This has some advantages. Since an engine has an optimal working point you have to use the transmission to try to keep the engine speed close enough to that point to sustain torque. If you tend to work too far from the optimal point using a fully mechanical transmission you need to introduce more gears. By using a torque converter you extend the interval on each gear that is close enough to the optimal engine working point, that is, a torque converter in a drive train means that fewer gears are needed. The torque converter also works as a shock absorber in the transmission. Using a torque converter has a drawback. The efficiency is not constant. The ratio between speed out and speed in is denoted nout /nin = ν. The highest efficiency is usually found at ν[0.6, 0.9] and when ν = 0 and ν = 1 the efficiency is 0. Figure 2.5 shows the efficiency of a torque converter at different ν:s. Depending on the machine the torque converter is to be used in, the optimal efficiency interval might occur at
2.3. Power distribution
10
Figure 2.5: Converter efficiency at different ν:s
lower or higher ν:s.
The transmission The transmission has four forward and four reverse gears. It is automatically controlled using several parameters, ν being the most important, for the most optimal gear shifting.
The drop box The drop box delivers power to the front and rear wheel axles.
Differential gears The differential gear distributes power from the drive shafts to the right and left wheels. They are equipped with differential locks that distributes more power to the wheel with the best traction.
Hub reduction The hub reduction is located at each end of the wheel axles, closest to the wheels. They gear down the shaft speeds one last time. This means reduced stress on propeller shafts and axle shafts.
Chapter 3
Hypothesis 3.1
The difficulties
We wish to describe a wheel loader given a specific task. The difficulties in this has two reasons. First of all, the performance of the loader is closely linked to the operators performance, and second, the dynamics of the wheel loader is very complex.
3.1.1
The operator
Given two experienced operators doing a specific task, a wheel loader can show two completely different results when it comes to fuel consumption and production. The reasons are several. An operator can, for instance, be more or less aggressive in using the accelerator pedal, that is, be a faster regulator. This results in higher fuel consumption as well as a slightly higher productivity. Another problem comes from the fact that the Volvo wheel loaders are built to be used at low engine speeds. An experienced driver used to other machines might not realize this and therefore operate the machine at a far to high engine speed and thus have a much higher fuel consumption. The resulting model will probably not need to be able to show the performance of a driver using a loader in a bad way. However it must be able to simulate the performance of both inexperienced and experienced, aggressive and calm operators and whatever other difference with impact we can find.
3.1.2
The dynamics of a wheel loader
In Volvo CE:s wheel loaders the engine delivers torque to both the transmission and to one or more hydraulic pumps (read section 2.3 for a more complete description). The dynamics of this is difficult to describe since, when filling bucket etc., the required torque and the required hydraulic pressure is by no means constant and at any given moment it is not possible to know if more or less power will be diverted to the hydraulics.
3.2. Purposed solution
12
However, during transport the bucket is more or less locked in position and therefore almost all engine torque is transferred to the transmission. This makes transportation far easier to model physically.
3.2
Purposed solution
The resulting model should be able to give an approximate value of the fuel volume consumed during a specific period of time doing a specific task. Due to the repeating pattern of most tasks for heavy machines, all that is need to do is to find the mean time and fuel consumption of a cycle of the specific task. The fact that the work will be done over a long time means that a large number of cycles will be performed. The differences that exists between different cycles will have little effect on the total work since n X V( ci Xi ) i=1 n X
D(
Xi )
= nσ 2 =
√
nσ,
(3.1) (3.2)
i=1
where V (x) and D(x) is variance and standard deviation of x and σ is the standard deviation of all Xi . This means that determining the variances of the stochastic variables will not be needed (Theorem 1 in appendix A). However, the same theorem tells us that n X E( Xi ) = n ∗ m, (3.3) i=1
where E(x) is the expectancy value of x and m is the expectancy value of all Xi . This means that an error in expectancy value will result in an error in the result of the same magnitude. This, in turn, means that the expectancy values need to be determined with a high precision for the calculation model to yield good results. Since different parts of a working cycle has different properties it is desirable to try do divide every cycle into smaller parts with similar properties. These parts are called phases and are linearly combinable.
3.2.1
Phases and parameters
The identification of phases have been done by looking at both physical reality as well as the operators intention. This means that if the operator feels that a new phase in the cycle is about to start, then it is. More on this in the phase descriptions below. The phases should also be as long as possible without loosing the advantage we gain by splitting up the cycles. The phases that were identified are filling bucket, reversing from bank, transport to receiver, emptying bucket, reversing from receiver and transport to bank. They are described below and in table 3.1 is a summary of possible parameters connected to each phase. The duration time and fuel consumption are of unknown distribution. The sum is however, in accordance to Theorem 2 in appendix A, of approximately normal distribution. Mean values will be used as expectation values.
13
Chapter 3. Hypothesis
Filling bucket This phase starts as the bucket hits the bank. As a result, the pressure on the plus side of the lifting cylinder goes up. It ends as a reversing gear is requested. This phase should be close to independent of the task since the time needed to fill a bucket is independent of where the material is going. However, the operator might be more inclined to overfill the bucket if he knows that transporting the material will take longer time. If this is the fact it need to be modelled. The size of the bucket as well as material and bank hardness are obvious parameters to take into account. The operators experience and driving style should also affect time and fuel consumption. Due to the heavy work of the hydraulics as well as the drive train, the difficulties mentioned in section 3.1.2 will make this phase very hard do model physically and is thus probably best modelled statistically.
Reversing from bank This phase begins as the operator requests a reverse gear and ends as a forward gear is requested. This means that the machine will be rolling backwards when the phase ends since the drive train is used for retardations. The operator however has more or less decided that this phase is over and has started to focus on the next phase. Parameters should include mass since a heavy machine requires more work to move. Also rolling resistance and the length of reversation should be taken into account. The operators experience might, and driving style should affect time and fuel consumption. During most of this phase the hydraulics as well as drive train should be working. The hydraulic work however should be fairly constant and this phase should be possible to model physically. An easier statistical model might however produce better results.
Transport to receiver This phase starts as forward gear is requested and ends as the bucket starts to empty, that is, as the pressure on the plus side in the lifting hydraulics cylinder starts to fall. If the transportation length is short this phase is about the same as reversing from bank since the bucket will constantly be going up. In a longer transportation however, the work of the hydraulics will have less effect and might in some cases be considered as not doing any work. If this is true we have eliminated one of the big difficulties in physical modelling of a working wheel loader and should therefore be able to construct a good physical model of this phase. In a longer transportation, ground structure also affects time and fuel consumption since it has a large effect on maximum speed. In the case of short as well as long transportation the main work done by the
3.2. Purposed solution
14
hydraulics is to lift the bucket to unloading height. This is a factor that might need to be taken into account.
Emptying bucket This phase starts as the bucket starts to empty and ends as the wheel loader starts to reverse, that is, the wheels rotates backwards. This means that the operator may use the drive train to retard without initiating a new phase. Parameters include bucket size and how easy it is to empty the bucket, that is, if the operator needs to be careful when emptying or just need to drop the load. This phase includes very little work which means that fuel consumption will be very low and the model more or less only has to return time. Due to the difficulties of quantifying the difficulty to empty the bucket a statistical model will probably yield the best results.
Reversing from receiver Negative velocity starts the phase and forward gear request ends it. This phase is very similar to reversing from bank. The difference is that almost no work is done by the hydraulics since the empty bucket is going down.
Transport to bank The phase starts as forward gear is requested and ends as the bucket hits the bank, i.e. the pressure on the plus side in the lifting cylinder goes up. Similar to transport to receiver with one difference, no lifting of the bucket means little hydraulic work. Modelling should follow the same pattern as transport to receiver.
15
Chapter 3. Hypothesis
Phase
Parameters
Filling bucket
Material, bank hardness, bucket, operator experience, operator style, length of transport phase
Reversing from bank
Total mass, rolling resistance, length of phase operator experience, operator style
Transport to receiver
Total mass, ground structure, rolling resistance, unloading height, operator experience, operator style
Emptying Bucket
Bucket size, how hard is it to empty
Reversing from receiver
Total mass, rolling resistance, length of phase operator experience, operator style
Transport to Bank
Total mass, ground structure, rolling resistance, operator experience, operator style
Table 3.1: Phases and possible variables
Chapter 4
Gathering data 4.1
Pretest thoughts
The tests should be done using a machine working in standard cycles. Interesting parameters are the effects of different operators, materials and ground conditions. As described in section 3.2, the phases in the purposed solution should be delimited by gear changes, pressure changes and directional changes. These parameters should therefore be logged. For easy identification of where in a working cycle we are, the angle of the lifting boom is also useful. To be able to look at work load and operator performance, engine speed, accelerator pedal and some value of delivered torque should also be logged. Also, since this thesis concerns fuel consumption, this value should not be forgotten. Since the cooling fan requires quite some torque when in use the work of the fan could be useful.
4.2
Test vehicle
For testing, a L180E was used. It was standard issue apart from these special features: • Tires: 800/65R29 XLD-65-1 • Swing out mud guards • Attachment brackets • Logging counterweight This resulted in a slightly heavier machine which should mean a slightly higher fuel consumption than a standard machine.
17
Chapter 4. Gathering data
Parameter Gear control signal Gear value Engine speed Accelerator pedal position Fuel consumption Fuel consumption Total fuel consumption Lifting boom position Lifting cylinder pressure Bucket accelerometer Cooling fan speed Signed drive shaft speed Trip
Source CAN Calculated CAN CAN CAN (requested) Calculated Calculated External sensor External sensor External sensor CAN External sensor Calculated
Unit rpm % mg/str kg/h kg % MPa m/s2 rpm rpm m
Table 4.1: Logged parameters in tests
4.3
Sensor set up
The on-board computer gather data from a lot of built in sensors on a wheel loader. These can be accessed through the CAN bus. Some parameters of interest are however not measured. This meant that extra sensors needed to be installed. These extra sensors where an angel sensor for the boom, a pressure sensor for the lifting cylinder, an accelerometer on the bucket and a new sensor for drive shaft speed. The last sensor was needed because of the need of signed velocity and the standard sensor only measures the number of revolutions. Fuel consumption is controlled by the on-board computer. The signal is however not usually accessible through CAN bus, though it is possible to request it to the bus. This needs to be done each time the on-board computer is restarted. Apart from logging computer data the wheel loader was fitted with a fuel measurement tube for reference purposes since the accuracy in using the CAN-bus for fuel consumption measurements had not been tested earlier. A fuel measurement tube is an approximately 2m long tube with a known diameter that can be used instead of the standard fuel tank. By placing the tube vertically the amount of fuel consumed can be read off with quite good precision. The dimensions of the tube were such that 56mm = 1l. To log the data from different sources a specialized computer, the eDaq, was used. It is basically a Linux computer with both digital and analogue interfaces for logging. The eDaq can be accessed through a TCP/IP interface. Since the eDaq is a computer it is also possible to let it do some calculations right away. The calculations the eDaq was set up to do were to transform fuel consumption from mg/stroke to kg/h as well as total fuel consumed, gear control signal transformed to actual gear and the drive shaft speed was integrated to total distance travelled. A summary of parameters logged is shown in table 4.1.
4.4. Performed tests
4.4
18
Performed tests
The tests were performed at at Volvo CE:s testing site in Eskilstuna. Three different operators performed a total of 21 different tests. Apart from several signal verification tests and transportation tests, 9 were short cycle hauler loading, 7 long cycle hauler loading, 6 were feeding a pocket and 1 were timber handling. The materials handled where timber, gravel, macadam, clay and blasted rock. Table 4.2 shows the tests done.
4.5
Obtaining data from test logs
The eDaq logs are in a raw data format and to access these Somat EASE was needed. Somat EASE is a program built for the after work of an experiment with viewing and some calculation tools. These tools where however slow in the version available at Volvo but EASE also offered the possibility to save the data in a MATLAB .mat-file. This means that by opening the files and storing them as MATLAB files, analysis in MATLAB was enabled. The test-logs where about 30 minutes long. Splitting these into cycles and phases was made semi automatically by a MATLAB function that was constructed. This uses the phase delimiters defined in chapter 3 to find the approximate phase shift positions and allows the user to correct this if needed. This correction is usually needed in phase shifts defined by hydraulic pressures. Because of different reasons not all tests and cycles were complete and therefore needed to be excluded from the analysis. These cycles are shown in table 4.3.
Table 4.2: The tests run
TIMBE CLAY KR1 KR2 VAG
L180VER1 L180VER2 L180VER3 L180VER4 L180BERG
08 19 19 19 19
08 SC2
07 STEEP 08 SC1
07 SC1
07 LC2
07 LC1
06 SC1
06 DHIGH
Short cycle hauler loading, old bucket Timber handling Short cycle hauler loading Feeding pocket Feeding pocket Flat transportation with bucket at different heights Fuel signal verification Fuel signal verification Fuel signal verification Fuel signal verification Short cycle hauler loading
Load and carry, unloading in hauler Load and carry, unloading in hauler Short cycle hauler loading, old bucket Driving up/down slope Short cycle hauler loading,
Test equipment testing Driving up/down slope, 2 runs/load Feeding pocket, BSS on, no bumps Feeding pocket, BSS off, bumps Feeding pocket, BSS on, bumps Feeding pocket, BSS off, no bumps Short cycle hauler loading Short cycle hauler loading Short cycle hauler loading, hauler 10m from dig phase Load and carry, unloading in hauler Short cycle hauler loading
04 05 05 05 05 05 05 05 05
1 BACKE KR1 KR2 KR3 KR4 SC1 SC2 SC3
Test description
Test run
– – – – Blasted rock
Clay Macadam Macadam –
Macadam
Macadam Gravel – Macadam
Gravel
Macadam
Gravel
Gravel
– – Macadam Macadam Macadam Macadam Macadam Macadam Macadam
Material Pettersson Pettersson Pettersson Pettersson Pettersson Pettersson Pettersson Pettersson Pettersson
Bjurman Bjurman Bjurman Bjurman
Mats Bohman Mats Bohman Mats Bohman Mats Bohman Stefan Pettersson
Esko Esko Esko Esko
Pauli Hansen Stefan Pettersson Esko Bjurman Esko Bjurman Stefan Pettersson
Stefan Pettersson
Pauli Hansen
Pauli Hansen
Esko Bjurman
Esko Bjurman
Stefan Stefan Stefan Stefan Stefan Stefan Stefan Stefan Stefan
Operator
– – – – 10.07
– 8.03 8.40 7.88 7.56 – 8.17 – – –
7.11
10.65
11.14 12.84 8.01
11.49
Mean production per cycle [tonnes] – – 7.85 Est. 8.0 – – 7.54 7.67 7.73
1-9 10-18 1-9 10-18
Auto throttle at 1505 rpm Auto throttle at 1105 rpm,Interrupted Auto throttle at 1105 rpm Mixed transportations
Faulty weighting system Faulty weighting system Total weights: 33.94, 34.48, 37.58
Cycle Cycle Cycle Cycle
Cycle 1-12 Cycle 13-30 (31-34 unloaded w/o receiver)
15 cycles normal (1-15) 15 cycles trying to maximize mass in bucket(16-30
–
– Weight: 38.1, 41.3, 30.2 – Weighting system breakdown after 4 cycles Faulty weighting system Faulty weighting system – – –
XX
19 Chapter 4. Gathering data
4.5. Obtaining data from test logs
Test run 05 SC1 05 SC1 06 SC1 07 LC1 07 SC1 08 SC1 08 SC2 19 CLAY 19 KR1 19 KR2
Faulty cycle 1-3 12 Last 1-3 4 last 12,15,18 3,6,9 4,7 all 1,2
Reason Faulty fuel signal Incomplete cycle Incomplete cycle Wrong unloading height Unloading without receiver Incomplete cycles Incomplete cycles Faulty cycles Faulty fuel signal Problem with transporter resulted in faulty cycles
Table 4.3: Discarded cycles with the reasons to discard them
20
Chapter 5
Resulting model The resulting model splits the task into phases and uses statistical or physical modelling depending on phase. Physical modelling has been tried when possible but if an easier statistical model has delivered better results this has been chosen.
5.1
Filling bucket phase modelling
The filling bucket phase is characterized by high work intensity, i.e. high fuel consumption. However when entering the filling bucket phase, intensity is low since too high torque when entering the bank might cause the bucket to get wedged in the bank. The work intensity is also low in the end since this phase shift is defined by a gear change and low engine speed is required in gear changes for comfort driving. For short, fuel consumption is initially low but is after some time at maximum. It goes however down just before the phase ends. This behaviour can be approximated with three strait lines such that = m(0 ≤ t < τ1 ) + m(τ1 ≤ t < τ2 ) + m(τ2 ≤ t ≤ Tf ill ) m ˙ max − m ˙ min m(0 ≤ t < τ1 ) = (m ˙ min + ) ∗ τ1 2 m(τ1 ≤ t < τ2 ) = (τ2 − τ1 ) ∗ m ˙ max m ˙ max − m ˙ min m(τ2 ≤ t ≤ Tf ill ) = (m ˙ min + ) ∗ (Tf ill − τ2 ) 2 mf ill
(5.1) (5.2) (5.3) (5.4)
where mf ill is total fuel consumed in filling bucket phase, Tf ill is total time of filling bucket phase and τ1 and τ2 is the time when full intensity is reached and left respectively. m ˙ max and m ˙ min are the fuel consumptions at maximum intensities and idling respectively. With a little work this transforms to mf ill = τ ∗
m ˙ max + m ˙ min + (Tf ill − τ ) ∗ (m ˙ max ) 2
(5.5)
where τ = τ1 + (Tf ill − τ2 ),
(5.6)
that is, the time the loader does not work at maximum intensity. How long τ and Tf ill are depends on the operator as well as material and bank hardness. An
5.2. Reverse from bank and reverse from receiver phases modelling
22
aggressive operator will have a shorter τ and a slightly shorter Tf ill . Materials harder to excavate will have larger Tf ill independent of operator.
5.2
Reverse from bank and reverse from receiver phases modelling
The mean times of these phases depend mostly on how long the driver need to reverse. In this model the user are given a statistical value corresponding to the chosen operator doing a 90◦ turn (45◦ in reversing from phases). In the case of a larger turn the user may change this as he or she see fit. The fuel consumptions, mf romBank and mf romReceiver ,are calculated as mf romBank mf romReceiver
= Tf romBank ∗ m ˙ max ∗ ¯i = Tf romReceiver ∗ m ˙ max ∗ ¯i
(5.7) (5.8)
where Tf romBank and Tf romReceiver are mean times of the phases, m ˙ max is fuel consumption at maximum intensity and ¯i is the mean intensity during the phase. ¯i depends on the operator as well as the rolling resistance. The user is given a value of ¯i corresponding to the chosen operator and a 3% rolling resistance. In the case of a larger resistance the user may change this as he or she see fit.
5.3
Emptying phase modelling
The emptying bucket phase is characterized by a very low fuel consumption due to the low amount of work that needs to be done. The duration time of the phase varies a lot depending on the receiver and, in the case with a hauler as receiver, how well the match is between the bucket and the receiver. If the bucket is well matched the difference in time between the first and the last bucket into the hauler will differ much. This is since the last bucket will have to be slowly dropped onto the hauler to avoid spill. Fuel consumption, mempty , is calculated as it is in the reverse from phases, that is mempty = Tempty ∗ m ˙ max ∗ ¯i.
(5.9)
Values of Tempty and ¯i are given corresponding to the chosen user and a well matched hauler taking 3 buckets. If the user see so fit he or she may change these values.
5.4
Transportation phases modelling
The modelling of the transport to receiver and transport to bank phases is made in two different ways. If the transportations are shorter than 10 meters a statistical model is used. If they are longer than 10 meters each, they are modelled physically. The reason for this division is that the precision of the physical modelling is not good enough on short transport legs. More on this in chapter 6.
23
5.4.1
Chapter 5. Resulting model
Statistical modelling
The statistical modelling is made using linear regression with rolling resistance as dependant variable. The differences in mass, transportation length and operator experience and operator style have too little effect on both time and fuel consumption to be possible to model in a good way. The resulting model becomes Ttransports mtransports
= 2 ∗ (3.9 + 0.2 ∗ µrolling ) = 2 ∗ (0.0255 + 0.0015 ∗ µrolling )
(5.10) (5.11)
where Ttransports and mtransports is total transport time per cycle and fuel consumption respectively, both to receiver and to bank, and µrolling is rolling resistance.
5.4.2
Physical modelling
The physical modelling of time uses engine and converter performance data to calculate traction force. This is used to calculate acceleration and thus velocity. This way the transport times Ttransporti i = 1, 2, . . . is calculated for each leg of a transport. Fuel consumption is calculated using the engine control data backwards. A delivered torque at a specific engine speed means that a specific mass fuel was injected. To do a little more thorough description lets go through the functions doing the work.
calcPower This function handles calculations involving the drive train. It uses model specific component data to calculate delivered torque and power at different converter-ν:s. Depending on how the function is called it can also return the traction force, the fuel consumption or required engine speed. The main part of this function is the calculation of delivered torque at different converter-ν:s. As described in section 2.3.1 Volvo’s wheel loaders are engine speed controlled and by changing position of the accelerator pedal you change the maximum engine speed allowed. Denote this engine speed ωmax . Since maximum torque is not always delivered at ωmax , the engine speed where maximum torque is delivered needs to be calculated. This is also done at different converter-ν:s since the efficiency of the torque converter differs greatly between different ν:s (more on this in section 2.3.4). We denote the wanted engine speeds ωνi First of all, the vector ω is constructed. Its lowest value is the lowest value of the engine performance data and its highest values is ωmax . Include also the reference engine speed for converter data (ωconverterRef ). Values of engine torque (Mengine ), hydraulics deduction (Mhydraulics ) and cooling fan deduction (Mcooling ) are interpolated to fit ω using tabulated data. By deducing charge pump torque (Mcharge ) and interpolated torques from Mengine , the usable propulsive torque (Mnet ) is acquired. At a specific νi , the tabulated converter data can be scaled using Mνi = ω02 ∗
MconverterRef , 2 ωconverterRef
(5.12)
5.5. Summing up the phases
24
where Mνi is scaled torque at the engine speed ω0 and MconverterRef is the tabulated converter data. ωνi is the ω0 at which the line defined by 5.12 intersects Mnet . By repeating this for all tabulated νi the required results are achieved. From these results it is easy to calculate traction forces etc.
fuelFromTorqueRPM The engine performance is controlled by an on-board computer. Depending on the required torque a specific fuel mass is injected and that torque is delivered. This fact can be used the other way as well. Given a specific delivered torque a specific mass fuel must have been injected and hence the fuel consumption is given. The torque delivered from the engine is actually the acquired torque from the combustion minus internal engine friction. This means that when we do these calculations we first of all need to add engine friction to the delivered torque. The calculated torque is then used to interpolate the injected fuel mass per stroke. Using the known engine speed this is the easily converted to [kg/s]
transport This method simulates the operator. It is presented with information about the transportation, divided into shorter legs, and what operator model to use, and uses this to model the operators usage of the accelerator pedal. The information required about the transport legs is initial engine speed, initial velocity, maximum velocity, rolling resistance, incline or decline, and mass. The operator is assumed to accelerate as fast as possible up to a point defined by operator type. At this point acceleration is lessened and the model tries to simulate an operator trying to keep the current speed. Also defined by the operator style is how fast he or she retards. This combined with final velocity decides when the operator releases the accelerator pedal and presses the brake.
5.5
Summing up the phases
There is a simple Graphical User Interface (GUI) that ties all these functions together. In this the user may change operator model, phase times and intensities if needed and set up the transportations. The GUI executes the calculations given the parameters set and the phase times and fuel consumptions are added up and the answers are returned to the user.
5.6
Parameters in effect
In chapter 3 a table of possible effecting parameters was set up (table 4.1). These have been considered and if the effect has been measurable and the parameter quantifiable the parameter has been used in the model. In the statistical models the effect comes from different tabulated values of parameters and in the physical model
25
Chapter 5. Resulting model
the parameters act as parameters in either operator model or machine performance calculations.
Chapter 6
Conclusions and future work 6.1
Model accuracy
No tests has been run apart from the ones used to build the model. This means that the precision of the model has only been tested on the data it was built from. These tests indicates that an experienced user will have no problem in setting the time parameters of the statistical calculations in such a way that the errors in the results are less than 5% of the measured values in both time and fuel consumption. As for the physical model of transportations, the error in time estimates are small (less than 5%). The estimated fuel consumption however is approximately 85% of the measured values (estimated values range from 80-90% of the measured values). One reason for this is believed to be that the efficiency of the drive train in the model is independent of engine speed. The truth is however that at higher engine speed the efficiency gets lower and thus more energy is needed to do the work needed. In spite of these errors the model is interesting since it should be independent of which wheel loader model is used. It is based on component data and machine weight and is easily changed when needed. Any statistical model needs new tests to be upgraded.
6.2
Requirements for good results
To get good results in modelling, the user is required to have good knowledge in fill times of different materials and bank hardnesses since these variables have such a large effect on the fill time and thus on the fuel consumption. The differences in time in different bank hardnesses of gravel can go from 6 seconds in loose bank all the way up to 10 seconds in a virgin bank. Since the time the machine is not working at maximum load will be approximately the same in both cases, the 4 extra seconds will have a large effect on fuel consumption (doubled in some cases) as well as cycle time (and thus productivity). Knowledge of times in the from bank and from receiver phases is also good but not required. The differences in these phases have shown to be small given a specific terrain and task.
27
6.3
Chapter 6. Conclusions and future work
Effects of boom suspension system
To examine the effect of BSS on production and fuel consumption two sets of runs were made with and without BSS. The sets where composed of one 30 minutes long load and carry run on even ground, and one 30 minutes long run with four 10 cm high bumps put out approximately 10 meters apart. No clear results could be drawn from the results. Visually however, no BSS on bumps meant that a fifth bump started building up after the last due to small amounts of macadam falling from the bucket at every run. This should result in higher need of maintenance on the transport leg. It is possible that, since the runs where only 30 minutes long, the operator could maintain a higher speed than possible in a regular work site. In a normal site the uncomfortable operating of no BSS might result in lower speeds over the bumps and accelerations when past them.
6.4
CAN-fuel signal verification
The fuel tube was used in several runs. When the machines auto throttle was set to a specific value the logged data were 99-101% of the actual value of the tube. In a pocket feeding task the value was 99.2% of the tube value. This leads us to the conclusion that logging fuel consumption by use of the CAN-bus results in sufficiently good values.
6.5
Future work
Wanted in the final model of calculations is that it is easily upgraded to be able to handle any change made to the wheel loaders. To upgrade the model of calculations as it is now, new tests needs to be done at any new release. This means that the new model needs to be fitted with eDaq and sensors and sent into work for a week. This is costly both in money and resources. However some further development would probably lessen these costs. Two paths are possible. Try to produce a better physical model or try to find some correlation between different model sizes. The physical model does not produce good enough results as it is now. A better model of losses (primarily in transmission) might change this. One other thing to look into is the operator model which is crude as it is now. Operator style in load and carry applications does not affect transportation time very much, however it affects fuel consumption. Needed for improvements is to better quantify different operator styles and how they affect acceleration, retardation and maximum speed in different conditions. It is possible that these steps might bring the model to the point at which it can be used in the reverse from bank and reverse from receiver phases as well as the transport to receiver and transport to bank phases. If this is achieved the project goal of an easily upgraded model is reached. The model will simulate filling bucket and emptying bucket statistically and everything else physically. The statistical models will be easy to keep upgraded since changes in expectation values of time and consumption in these phases is easy to measure. The physical models will be upgraded through new component data and will only
6.5. Future work
28
require some verification testing. The other path will result in a cruder model that is harder to upgrade. It will however not require any further theoretical work, only extensive testing. No matter which path is chosen, more tests in different material and bank hardnesses with different operator styles should be done. This would produce a more complete material knowledge in the model and would lessen the need of user knowledge in phase times. It would also help in quantifying operator styles.
Bibliography [1] Malmberg, Carl Einar et al. Terr¨ angmaskinen 1. Gummessons Tryckeri AB, Falk¨ oping 1993. ISBN 91-07614-083-0. [2] V¨ annman, Kerstin Matematisk statistik. 2nd edition. Studentliteratur, Lund 2002. ISBN 91-44-01690-5. [3] Blom, Gunnar. Sannolikhetsteori med till¨ ampningar A. 2nd edition. Studentlitteratur, Lund 1984. ISBN 91-44-04372-4. [4] Montgomery, Douglas C et al. Applied statistics and probability for engineers 2nd edition. John Wiley & Sons, Inc., New York 1999 ISBN 0-471-17027-5 [5] Filla, Reno Operator and Machine Models for Dynamic Simulation of Construction Machinery. Licentiate thesis, Department of Mechanical Engineering, Link¨ opings universitet, Link¨oping, Sweden, September 16, 2005. Permanent link: http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-4092
Appendix A
Statistical background Theorem 1 (Linear combination of locatives variables). For all stochastic variables X1 , X2 , ..., Xn we have n X E( ci Xi ) i=1 n X
V(
ci Xi )
n X
=
i=1 n X
=
i=1
ci E(Xi )
(A.1)
ci V (Xi ) + 2
i=1
X
ci cj C(Xi , Xj ).
(A.2)
i =
√
nσ.
(A.5) (A.6)
i=1
Theorem 2 (Central Limit Theorem). If Xi i = 1, 2, .. is an infinite series of equally distributed stochastic variables with an expected value of m and a standard deviation of σ > 0 and if Yn = X1 + X2 + ... + Xn then P (a
