torque

angle

2 | SIMPACK News | November 2009

CUSTOMER APPLICATION | Franz Froschhammer, BMW Group; Marcus Schittenhelm, Rainer Keppler, SIMPACK AG

BMW High-Dynamic Engine Test Benches using SIMPACK Real-Time Models

The term 'component-based' refers to models which include individual components. For example, suspensions are modeled with the suspension arms, bushings, dampings, etc., as opposed to using quasi-static look-up tables to describe the suspension's characteristics. The term 'high-dynamic' refers, in this context, to simulation of the longitudinal dynamics of the vehicle. The frequency range of up to approx. 15 Hz, in which the typical vehicle jerking oscillations after load changes occur, can be observed (see Fig. 1). In order to

apply vehicle handling behaviour directly to the engine test bench, an exact simulation of the real vehicle behaviour has to be used [1].The necessary quality of reproduction in the relatively high frequency range for real-time

applications, is achieved by using detailed, component-based SIMPACK real-time models

on the test bench. The parameterisation of these real-time models takes place exclusively with original design data and not measurement or MBS pre-processing data.

As far as time and quality are concerned, the only way to meet the increasing development costs of modern combustion engines is by using the most modern design methods, a suitable infrastructure, and optimised procedures.As a result of the continually growing pressure on costs, and the requirement to reduce the number of both test drives and prototype vehicles that go along with it, the necessity to shift testing from the vehicle to the test bench is becoming increasingly important. For these tasks, high-dynamic engine test benches in combination with component-based SIMPACK real-time models are used by the BMW Group.

Fig. 1: Acceleration after load changes: the test bench helps to find the optimum settings.

“...component-based SIMPACK real-time models are used by the BMW Group”

SIMPACK News | November 2009 | 3

Franz Froschhammer, BMW Group; Marcus Schittenhelm, Rainer Keppler, SIMPACK AG | CUSTOMER APPLICATION

Implementation of these requirements is achieved by generating the necessary real-time models (‘Online’) with the help of the SIMPACK code export functionality. The highly detailed SIMPACK MBS-models (‘Offline’), that BMW already uses in the earliest development phase, are used as the basis for the real-time models. The significantly increased detail of the real-time models allows their usage to be moved to earlier phases of the development process. In addition, there is an opportunity to carry out detailed, component-based variant calculations on the test bench. All of this contributes to a reduction in development costs while providing better results in the early phases of development.

BMW APPLICATION AND TEST BENCH SCENARIOThe high-dynamic engine test bench consists of a combustion engine with a control unit (DME) which is connected via a connecting shaft to an electric motor. Depending on the vehicle model, a transmission control unit (EGS) can be added.The connecting shaft is specifically designed for every engine type with regard to stiffness and damping, in order to adjust

the mechanical natural frequency of the test bench to the desired value.Since only the combustion engine together with the control unit(s) are actually present

Fig. 2: High-dynamic engine test bench system configuration at BMW

on the test bench, all other components of the vehicle must be reproduced with the aid of the simulation model. The reaction of the vehicle to the stimulus from the combustion engine is represented by the tightly coupled electric motor which receives its desired value directly from a vehicle model. A detailed, component-based SIMPACK real-time model provides this important building block in the test bench simulation scenario (see Fig. 2).In order to make full use of the expertise already present in the BMW Group concerning the integration of real transmission control units, BMW's transmission model is included. Thus, the interface between BMW's own test bench environment and the SIMPACK vehicle model can be found at the output of the transmission model. At this connection, the SIMPACK model provides the relevant information on the angle of rotation as an input for the test bench. The input for the SIMPACK model is the corresponding torque from the transmission.

Fig. 3: 3D side view of a complete BMW SIMPACK real-time model

Fig. 4: Covered frequency range of SIMPACK real-time models on BMW high-dynamic engine test benches

4 | SIMPACK News | November 2009

CUSTOMER APPLICATION | Franz Froschhammer, BMW Group; Marcus Schittenhelm, Rainer Keppler, SIMPACK AG

SIMPACK REAL-TIME SIMULATION MODELSIn order to meet the requirements for model validity in the relevant frequency range (Fig. 4) and the requirements for component-based calibration, it was necessary to use corresponding detailed MBS models. The dynamic oscillation behaviour in a vehicle due to a load change is a complex interplay of drivetrain and axle oscillation. Therefore, the models must be able to provide a detailed representation of the spatial and dynamic coupling effects between the drivetrain, 3D wheel suspension oscillations and axle carrier or differential housing oscillations. This means, in turn, that the usual approach to real-time driving dynamics applications quasi-static modelling approaches (e.g. using look-up-tables from compliance measurements) with regard to axle oscillations cannot be used. A completely component-based MBS modelling approach was chosen in which all components relevant to oscillation are represented as separate MBS elements. Currently used SIMPACK real-time models (consisting of bodies, joints, and force elements) have more than 100 degrees of freedom (see Fig. 3 and Fig. 5). This approach offers the following important advantages:

• A precise model which includes a highfrequency range content relevant for driving dynamics applications.

• Model parameterisation based purelyon design data. Physical and/or virtual measurement data is not required (i.e. no restrictions with regards to alternative drive types such as RWD, FWD, and AWD).

• Use of an already existing databasecreated for SIMPACK "Offline" oscillation comfort models.

Fig. 5: 3D rear view of a complete BMW SIMPACK real-time model

Fig. 6: Principle 2D view of a SIMPACK real-time model topology at BMW

• Parametric sensitivity analyses of theinfluence of individual vehicle hardware characteristics are possible directly on the test bench without additional MBS pre-processing (e.g. axle component masses, drivetrain inertia, rubber bearing characteristics).

An example of the complexity of the topology of an applied real-time model can be seen in Fig. 6.The challenge of being able to use such detailed MBS models in a real-time system is solved by SIMPACK Code Export. This makes it possible to export SIMPACK "Offline" models of any complexity, generated from MBS library elements as Fortran or C-Code to the corresponding real-time test bench environment, as "Online" models. In order to ensure that necessary calculation speeds and solver stability requirements needed for the real-time system are met, unique SIMPACK methods (SIMPACK minimum coordinate approach, SIMPACK ODE generation of equations of motion), and newly developed real-time solvers for this sort of application are used.

THE IMPLEMENTED PROCESS FOR MODEL GENERATION AND CALIBRATIONApart from model generation and ensuring numerical stability, the model calibration process

for productive and reliable application of a simulation-based develop-ment solution

is possibly the greatest and most expensive challenge. The earlier the simulation solution is to be used within the development process, where less vehicle test data is available for comparisons, the more important simulation-based development becomes. With this application, a realistic simulation is achieved by basing the entire model generation and calibration process on SIMPACK Offline models which are used for whole vehicle oscillation design. These models are already available at the BMW Group from a very early development

phase. In this way, considerable synergies are exploited by using the same model and parameter database for both "Offline" and "Online" models. Changes of the model parameters can be transferred directly from the "Offline" to the "Online" model without additional pre-processing steps or redundant data entry (see Fig. 7).

ExAMPLE OF AN APPLICATION ON THE HIGH-DyNAMIC ENGINE TEST BENCH: POSITIvE LOAD CHANGE FROM DyNAMIC IMPULSEThe detailed representation of the vehicle longitudinal dynamics allows, for example,

load changes to be applied. While the effects on the drive behaviour are directly perceptible in the vehicle, the calibration engineer

assesses them on the test bench with measurement processes. The criteria for the design are determined by the objectives, such as measuring the number of jerking oscillations and/or amplitude values [1]. A comparison between the vehicle and the test bench is shown in Fig. 8.

“...considerable synergies are exploited by using the same

model and parameter database for both "Offline"

and "Online" models”

“Currently used SIMPACK real-time models have more than 100 degrees

of freedom”

SIMPACK News | November 2009 | 5

Franz Froschhammer, BMW Group; Marcus Schittenhelm, Rainer Keppler, SIMPACK AG | CUSTOMER APPLICATION

Fig. 7: SIMPACK "Offline" and "Online" model generation and parameter supply process at BMW

Fig. 8: Calibration of positive load change due to a dynamic impulse: vehicle-test bench comparison

SIMPACK "Online" application: High-dynamic engine test bench

SIMPACK "Offline" application: e.g.hydro-pulse test rig, bad road simulation etc.

SUMMARy AND FORECASTThe method described for engine calibration on the high-dynamic engine test bench using detailed, component-based SIMPACK real-time models has been implemented successfully. Moving those tasks to an earlier phase in the development process, at which time no test vehicles are available, offers a considerable potential for early achievement of calibration targets and cost reduction.In addition, the opportunity to carry out sensitivity analyses of vehicle hardware characteristics opens up many new possibilities for optimising future calibration tasks on high-dynamic engine test benches.

REFERENCES[1] Franz Froschhammer, Detlef Mathiak, Friedrich Rabenstein, 'Hochdynamische Prüfstände — Ein

Werkzeug für die Insta-tionärapplikation',ATZ/MTZ Konferenz — Motor, Motorenentwick-lung auf dynamischen

Prüfständen, Wiesbaden, 23.– 24.11.2006

“...opens up many new possibilities for optimising future calibration tasks...”