DE VELO PMENT ME A SURING TECHNIQUES
© John Deere
DLG PowerMix on the Dynamometer
John Deere implemented the end customer test DLG
MOTIVATION
PowerMix to its dynamometer in the development centre
The DLG PowerMix, a testing process of the German Agricultural Society (DLG), has become a de facto standard for testing the fuel consumption of tractors. The test is practice-oriented because it determines the specific operating costs (diesel, urea) in the most common agricultural workloads. Traction, PTO and hydraulics are loaded realistically as on the field or during transport drives [1], TABLE 1. As fuel makes up 30 to 50 % of the overall operation costs of a tractor, farmers use the test results to make their buying decisions. An app [2] enables users to compare the tractors tested according to the specific requirements of their farms, listed in €/kWh. Agricultural magazines present the test results in their reports. At John Deere, the tar-
in Mannheim. A test solution from AVL turns the test bed into a powerful X-in-the-loop test and development environment. The engineers can now make accurate predictions about the specific fuel consumption in work and transportation cycles long before the tractors are finally constructed. This innovative model-based test method, which is embedded into a robust, industrialised process, helps to save time and money. It also provides important insights into the dependencies of the subsystems in the overall vehicle under real-world customer operating conditions. 26
AUTHORS
geted PowerMix results are part of the specifications. DLG POWERMIX AS PART OF THE DEVELOPMENT PROCESS
Claudia Pieke, M. Sc. is Test Engineer at John Deere GmbH & Co. KG in Mannheim (Germany).
Dipl.-Ing. (FH) Waldemar Stark is Test Engineer at John Deere GmbH & Co. KG in Mannheim (Germany).
Dipl.-Ing. Dr. Felix Pfister is Business Development Manager Business Segment Virtual Testing at AVL List GmbH in Graz (Austria).
Dr.-Ing. Christian Schyr is Project Manager Virtual Testing Solutions at AVL Deutschland GmbH in Bietigheim-Bissingen (Germany).
Almost all European tractor manufacturers bring their tractors to the DLG test centre in the German city Groß-Umstadt to test them in the PowerMix before product release. This procedure is expensive and time-consuming. It also poses a risk to the development process, because the tests can only be conducted with a fully-functional tractor at the end of the development process: at this development stage, it is normally too late for systematic corrections. In order to predict the consumption earlier and use the findings for series development, John Deere in Mannheim reproduces the DLG consumption tests using model-based methods on the powertrain test rig. The environmental conditions of the test rig can be controlled exactly, which is necessary to identify savings potential and increase fuel efficiency through calibration of hardware and software parameters. In general terms, model-based testing means embedding a real unit under test
(UUT) into a virtual overall vehicle (virtual integration of real prototypes) and test driving it in simulated manoeuvres [4], FIGURE 1. This way, a predicted test report can be generated on the test bed. This front-loading of tests resulted in increased fuel efficiency for the newest tractor generations. Despite higher complexity, development times were reduced. This X-in-the-loop approach is described below. DYNAMOMETER AS MECHATRONIC DEVELOPMENT ENVIRONMENT
In recent decades, the technical device tractor has developed from a simple, pure mechanical system to a complex, networked mechatronic system. This growing complexity imposes higher requirements on the testing technology. Test bed and UUT should always be on the same technological level. The test bed in the laboratory must stimulate the same interfaces that the system tractor will use in later real-world communication with its environment. These interfaces affect the power flow (mechanical, thermal, electrical), the mass flow and the flow of information (sensors, ECU communication). In model-based testing,
No.
Load type
Tractor load
Working process
01
Drawbar work
100 %
Heavy ploughing
02
Drawbar work
60 %
Medium heavy ploughing (light soils, small plough)
03
Drawbar work
100 %
Deep cultivating
04
Drawbar work
60 %
Medium heavy cultivating (for example stubble cultivating)
05
Drawbar and PTO work
100 %
Heavy work with rotary harrow (for example deep depth)
06
Drawbar and PTO work
70 %
Medium heavy work with power harrow
07
Drawbar and PTO work
40 %
Light work with power harrow (for example at shallow depth, light soils)
08
Drawbar and PTO work
100 %
Heavy mowing (for example first mowing or with conditioner)
09
Drawbar and PTO work
70 %
Medium heavy mowing (for example second mowing or small crops)
10
Drawbar and PTO work
40 %
Light mowing (for example only front mower)
11
Drawbar, PTO and hydraulic work
100 %
Manure spreading
12
Drawbar, PTO and hydraulic work
60 %
Baling
13
Transport work
100 %
Heavy transport work (for example uphill transport work)
14
Transport work
25 %
Light transport work (for example flat transport work)
TABLE 1 The DLG PowerMix measures the real-life consumption of tractors under partial and full load conditions [3] (© John Deere) ATZ offhighway worldwide
02|2017
27
DE VELO PMENT ME A SURING TECHNIQUES
bed interface models that establish a robust coupling between the real component on the test bed and the virtual vehicle and virtual environment in the simulation [5, 6]. This allows for a customer-oriented evaluation of virtual/real vehicle components in the context of the overall drivetrain and vehicle. FROM REPLICATION TO MODELAND MANOEUVRE-BASED TESTING
FIGURE 1 A virtual tractor with an integrated real drivetrain moves on the digitised DLG transport section – the vehicle is driven by a virtual or a real driver (© John Deere)
a “real-life” simulation environment feeds these flows. FIGURE 2 shows the all-wheel powertrain test bed in the plant in Mannheim together with a tractor of the 6-series. The engineers in Mannheim are responsible for developing and testing this series. Each year round about 30,000 green tractors with their characteristic
yellow rims leave production in Mannheim to be shipped to destinations all over the world. The test solution AVL InMotion turns the test bed into a powerful X-in-the-loop test and development environment. The test solution expands the established software IPG CarMaker/TruckMaker to include advanced and sophisticated test
From 2010 to 2015, John Deere generated the DLG PowerMix by replicating timebased load profiles [7]. In principle, this procedure was successful, but it was not efficient and flexible enough. It was replaced by model-based testing for the following reasons: – The calculated test run was only valid for one specific tractor model. If the tractor was modified (keyword: variant diversity), the load profile had to be recalculated using simplified inverse simulation models. This procedure was time-consuming and did not produce the required prediction accuracy.
FIGURE 2 The dynamometer stimulates the interfaces between the real, mechatronic system tractor and the simulated environment (© John Deere)
28
FIGURE 3 Test bed configuration: The existing chassis dynamometer was expanded to include a virtual test environment (© John Deere)
– Transport workloads, which have been part of the PowerMix test programme since 2014, could not be mapped with satisfactory quality. – The procedure was not flexible and forward-thinking. The set point-based tests could not map the diversity of (future) customer applications. After a comprehensive, one-year evaluation and implementation phase, John Deere shifted to the model- and manoeuvre-based X-in-the-loop test process. This sophisticated method has been successfully integrated into the test bed operation, because, from the beginning, equal emphasis has been placed on the three main factors: product, process and people. PRODUCT
As mentioned above and presented in [8], AVL InMotion was integrated into the existing landscape of test beds as a central tool, FIGURE 3. Through actors (for example dynamometer) and sensors, the test solution merges the real powertrain prototype, which is mounted on the test bed, with the virtual prototype of the tractor trailer combination into a “system of systems under test.” This virtual prototype includes the real-time, standardised, non-linear MBS models of TruckMaker with axles, steering, tyres, ATZ offhighway worldwide
02|2017
braking system, hydraulics, etc. The simulation environment is controlled remotely by the test automation system AVL Puma Open through a powerful TCP/IP interface. The 3-D online visualisation with a connection to Google Earth gives an overview on the current progress of the tests, FIGURE 4. For future projects, AVL InMotion includes prepared interfaces for Testbed.Connect and Model.Connect. Unlike the replication of load profiles, the tire loads are not calculated in advance. Instead, the PowerMix manoeuvres are defined and the driving physics, e.g. tyre physics (rolling resistance, slip, etc.) and momentum of the trailer are calculated online and in real time. Test description and test object are clearly separated. The advantage: once implemented, the manoeuvres can be used for different vehicles. The DLG PowerMix test catalogue, including the digitised tracks for transport drives, was implemented by AVL in close cooperation with John Deere as AVL TestWare, which is independent of specific models or test bed types. PROCESS
The delivery of real powertrain prototypes to the test factories has been an established process or decades. Up to
now, this only applied to a certain extent to virtual prototypes (models). During the implementation of the new methodology, great importance was placed on installing a robust process for the generation of virtual prototypes with properly documented development levels according to the real prototypes. In other words: model-based testing changes processes. In fact, these changes must be implemented in the tough everyday reality of a test factory. Only this way can the product’s potential for efficiency gains be explored fully. PEOPLE
The two pillars product and process are necessary but not sufficient. People is the first and most important pillar. To get people on board, they have to be mobilised for the new tasks. Key to the project’s success has been the early involvement of the test bed staff: during the product and process implementation, the employees were trained on the job for their new tasks. They thus participated in the project’s success right from the start. In other words: people have to identify with the approach of modelbased testing. If they do not, the project risks failure – although the industrialised products and robust processes
29
DE VELO PMENT ME A SURING TECHNIQUES
enclose the complexity of simulation and keep it away from the test bed staff. FORWARD -LOOKING TEST APPROACH FOR CUSTOMERORIENTED DEVELOPMENT
FIGURE 4 Comparison of model-based test results with DLG results in PowerMix cycle 02 (plow 60 %) (© John Deere)
The test procedure predicts the DLG PowerMix results with ±3 % accuracy, FIGURE 4 and FIGURE 5. This accuracy lies, and this is actually the most important information, within the standard deviation of real-world tests. In addition, the scatter band of the results (measurement noise) on the test bed is smaller by a factor of 2 to 10 than in real-world operation. Only this way is it possible to achieve the necessary discriminability to evaluate individual effects. Looking back on the one-year implementation and two-year productive phase, the following advantages of the new testing procedure can be noted: – Expensive, time-consuming and weather-dependent on-site tests can be reduced to a minimum (reduction by approximately a factor of 2). – Comprehensive fuel measurements in customer use cases show significant savings potential. – Fuel consumption results can be predicted accurately one or two iteration loops earlier than before. This means that results are obtained six to twelve months earlier. The risk for failing to meet the specification requirements is reduced. – If variant diversity increases, critical boundary samples can be identified and tested in the DLG tests using the model-driven approach. – The test solution is open, flexible, and extendable; not only for the planned DLG PowerMix 2.0. Model-based testing is future-proof and will advance developments in the area of electrification and automation. SUMMARY
FIGURE 5 Comparison of model-based test results with DLG results in PowerMix cycle 13 (transport, part run number 5, uphill) (© John Deere)
30
John Deere is well prepared for the challenges of a mechatronic test bed of the future. The new test approach endorses to the companies claim to optimise tractors not for artificial cycles, but for the real demands of their end customers. The demands placed on tractors are more diverse than for any other vehicle. With model- and manoeuvre-based testing, product performance in customer
applications can be increased significantly. In addition, the development cycles are shortened, something that is indispensable for success on a highly competitive market. REFERENCES [1] Degrell, O.; Feuerstein, T.: DLG-PowerMix – Ein praxisorientierter Traktorentest. In: 61. Tagung Landtechnik. Hannover, 7-8 November 2003, pp. 339-345
ATZ offhighway worldwide
02|2017
[2] N. N.: www.dlg-test.de/powermixapp [3] N. N.: DLG-PowerMix um Transportfahrten erweitert. DLG Press Release, Frankfurt am Main, 20 May 2014 [4] Paulweber, M.; Lebert, K.: Mess- und Prüfstandstechnik. Wiesbaden: Springer, 2014, pp. 321-337 [5] Pfister, F.; Schyr, C.; Le Rhun, F.: Total Energy Efficiency Testing. The Chassis Dynamometer as a Mechatronic Development Platform. In: ATZ 111 (2009), No. 11, pp. 45-48 [6] Riedler, A.; Abt, R.; Pfister, F.; Rieger, H.: Prüftechnik von morgen schon heute – Einbettung realer KTM Motoren in eine virtuell
simulierte Welt. In: Proceedings of International Symposium on Development Methodology 2009, pp. 198-209 [7] Back, P.; Hodel, B.; Pirro, P.; Stark, W.: Replication of the DLG PowerMix Tractor Fuel Consumption Test in a Laboratory Environment. In: Solutions for Intelligent and Sustainable Farming, AgEng – Land.Technik, Internationale Tagung Landtechnik, Düsseldorf, 2011, pp. 209-215 [8] Pieke, C.; Stark, W.: PowerMix goes Rig. New Methodologies for Tractor Efficiency Optimization. In: AVL Product Development in Motion 2016, Gothenburg, 2016
31