How Is Simulation Being Used In Automotive Design?

Louise Davis

Instead of designing motor components one after the other, Porsche and Altair have teamed up to use simulation to do it all at once

The rate of which electric hybrid and full-electric vehicles are reaching the market and being developed has increased considerably over the past years and the requirements on the e-motors being designed for new cars are increasing rapidly as well. The goal is to develop better motors within tighter time and cost schedules. 

Simultaneously, the technical requirements on the motors are increasing rapidly, both in terms of level and bandwidth of requirements. 

Today, an e-motor cannot just be developed looking at the motor as an isolated unit; it must be assured that tight requirements concerning the integration into both the complete electric or hybrid drivetrain system and tight requirements concerning perceived quality are fulfilled. Thus, it is a necessity to develop the e-motor not in isolation but as a system to fit optimally with other components and systems. Noise and power consumption are two of such integration challenges.

What Is Simulation Driven Design?

FE and other simulation methodologies have traditionally been used very successfully to verify designs and design directions. Today, FE and especially numerical optimisation is increasingly used to support and drive the design process, i.e. optimisation is used to help the design team find best alternatives, executing sensitivity studies, performing trade-offs between different design alternatives, and so on. This design strategy is often denoted ‘simulation-driven design’. 

It’s especially beneficial where the design is less intuitive because of high design complexity and/or complexity of loads and targets for the design. Products and designs that experience requirements from several different types of physics and attribute disciplines are especially suited for using simulation-driven design since it quickly becomes impossible to comprehend the relationships between design change and change in behaviour using traditional design methods.

Multidisciplinary and multiphysics optimisation methodologies make it possible to design an e-motor for multiple, completely different design requirements simultaneously, thus avoiding a serial development strategy, where a larger number of design iterations are necessary to fulfil all requirements and unfavourable design compromises need to be accepted. 

Multiphysics and multidisciplinary optimisation, however, need efficient processes to be executed within the narrow constraints and time limitations of a live product development. The processes need to be integrated with all departments in the e-motor development.

How Are Porsche Testing Simulation?

A baseline design is used as a starting point for the optimisation. A design space is then created by defining variables (design variables, DVs) that influence the design. 

In this study shape variables, which influence the size and position of the magnets, are used to create the design space. Then, the essential responses are selected. 

Depending on the choice of responses, one or more solvers must be used to perform one or more simulations to yield the necessary responses. In certain cases, co-simulation of solvers is necessary to resolve responses depending on each other, i.e. multiphysics situations, such as where situations with time-dependent output from one solver is needed to solve responses in another solver and vice versa. 

Before launching the study, general study parameters must be defined and how the optimisation should be executed. If metamodel-based optimisation is chosen, response surfaces of all responses are created based on the samples from the DoE (Design of Experiments). 

Optimisation and design exploration can then be carried out using these response surfaces. A strength of DoE-based optimisation is clearly the ability to use the data to answer a large number of different questions and to play through numerous design scenarios.

What Is The Porsche/Altair 3 Step Design Process?

Porsche is developing high-performance e-motors with high requirements on key performance data such as power, torque and speed. Porsche and Altair agreed on applying a three-step, optimisation-driven design process to develop a concept.

The first phase supports the development of a baseline combination of stator and rotor concepts focusing on magnet configuration. For each magnet configuration, design optimisation is executed to derive an optimal set of design parameters.

In the second phase, the design scope is extended to include other important physics to be considered during the e-motor design process. In addition to phase one, heat transfer, structural strength and demagnetisation responses are added to the design problem.

The third phase is focused on looking at the e-motor in its environment and thus including other parts of the drivetrain. As a first step, the inverter is added to generate more realistic currents into the e-motor design process. Later, the systems approach will be used to calculate efficiencies and temperature development for complete drive cycles.

The details

The first phase concerns the task of finding the right starting point for the multiphysics design process. Altair’s FluxMotor was chosen for this task. Based on a classical rotor topology, different winding configurations were investigated with respect to maximum torque and power for one working point close to the base point. When the preferred winding configuration has been found, the next task is to find the best matching rotor configuration based on the design requirements stated for the e-motor to be developed. In this project, four competing rotor designs were investigated and compared. 

In FluxMotor different test scenarios are available to analyse a motor concept and for the requirements, the ‘efficiency map test’ was chosen to compare the four different topologies. From this test, both the base point and the max point data could be extracted and used for the comparison of the designs. Four different topologies were tested.

To satisfy different requirements coming from different physics, a strategy was chosen to use a multidisciplinary optimisation process, in which several computations using different tools were used. The different tools and simulations where necessary to calculate all requested responses for the multiphysics design optimisation problem. The simulation types and the working points were chosen such that all responses could be extracted using a minimum of calculation effort.

To study the mentioned working points, FluxMotor was first used to extract the main characteristics of the motor, such as speed, current rms and control angle values. These values being known, FE-based tools could be used to accurately calculate all motor characteristics including iron losses, efficiency, etc. Flux 2D was the tool used to completely derive the e-motor behaviour at three different working points. 

The input data for all these simulations includes material properties, the electric properties (resistance and current supply) and the speed of the motor. After having reached steady-state for these simulations, the following output data could be extracted: torque, torque ripples, losses and efficiency.

For the last simulation, 100kW at max speed test, a dedicated analysis was done for extracting losses in each region (iron losses in rotor and stator, eddy current losses in magnet, Joule losses in the coil). These are later used to feed the thermal simulation.

The cooling is done through a water jacket on the outside of the stator. Convection and radiation are accounted for through boundary conditions. The test case prescribes to stay at maximum speed for two hours. The goal is to check that there is no risk of overheating the coils. The losses determined from the previous test are used as input for this test. Finally, the temperature is determined as a function of time, to test the coil winding temperature at the last time step.

When designing e-motors, it must be assured that the risk of magnet demagnetisation is minimised. Porsche uses a short-circuit test at the base point to address this issue. Based on such a simulation, a specific feature and procedure in the software is applied to compute the remnant flux density at the end of the computation. We can then extract a percentage of the magnet that is demagnetised.

The challenge is to get the highest value of current after short-circuiting. A parametric analysis has shown when to start the short-circuit.

How Is Mechanical Stress Avoided?

Mechanical stresses must be constrained to be kept below a specific level to assure mechanical integrity. The stress occurs mainly due to rotational forces at high speed. The starting point is a STEP file generated as a result of the Flux 2D load cases. Based on the geometry information in the file, an FE mesh is created, and all mechanical properties are automatically set in a batch process within Altair HyperMesh. The simulation to evaluate stresses is executed in Altair OptiStruct. The maximum stress values are finally extracted. At this point the focus lies on tensile stresses since they are considered more critical in comparison to compression stresses.

The complete study with all simulations was setup in Altair HyperStudy. Total run time to extract all responses for one single design was 29 minutes. A DoE with 358 runs was executed to cover the design space. The total run time was 17.45 hours running 15 jobs in parallel. Following the DoE, optimisation and design exploration could be performed. Such optimisations and design explorations can be executed on a subspace of simulation types and responses and on the complete thing. The optimisation problem can be formulated as a single or multi-objective optimisation problem.

Using the DoE data, studies can be executed on a subset of variables or on the complete problem. Studies can be performed concerning driving design variables, sensitivities and trade-offs between different design objectives and constraint settings.

Adding The Power Inverter

In the final phase, how to improve the design of the motor is considered by adding the power electronics which improve the accuracy of the inputs driving the machine.

The design process up to this point has assumed the inputs to the e-motor are idealised (i.e. purely first harmonic) sinusoidal inputs to the three phases. However, the actual system supplies input voltages based on modern power electronics and pulse width modulation (PWM) techniques to approximate the desired driving voltage from the control algorithms in the system. 

This particular system includes a two-level inverter along with a current and speed controller cascaded to drive the logic of the transistors with space vector pulse width modulation.

PWM methods create higher order harmonic content of the electrical inputs which can degrade certain aspects of the performance off the e-motor, such as losses and torque ripple in the machine. The losses affect efficiency and the thermal behaviour, and the torque ripple cause speed pulsations and NVH problems, and thus simulating the inverter and dependent systems in an important aspect to getting toward an optimal design.

Electromagnetic losses also contribute to the thermal behaviour, which is very important for design: consideration for critical components such as coils and magnets are important to capture accurately and this helps to improve the accuracy for the cooling system design as well. 

In this way we can get a more accurate result that leads to more confidence that the e-motor has been designed optimally compared to using the prior design process up to this point. Altair Flux does have electric circuit building capabilities, which is useful for simple circuits; however, to achieve more accuracy for the control system, we will use a different simulation tool that is better suited for this task and integrate this into the design process. 

Building this system will be accomplished in Altair Activate, a multidomain system simulation environment that will allow us to model the complex inputs not only for the power electronics, but also the detailed controls algorithms, in this case using space vector pulse width modulation (SVPWM).

Summary

In the project new methodology and new processes have been developed to successfully execute all tasks and phases. Since the execution in the optimisation engine (HyperStudy) requires that all steps can be executed automatically, effort has been spent on designing batch scripts which support the automatic execution of the process.

The processes have been developed on the basis of the Altair HyperWorks suite of tools which have open APIs and can be executed in batches.

This shows the potential of using simulation-driven design. As soon as the optimisation process has been successfully setup, it can deliver significant information about design directions, sensitivities and the consequences of design choices made during the development.

The studies performed so far show that an initial design can be improved considerably when run through the design optimisation process.

The process described here has not yet reached the final state. Porsche and Altair are working on improving it and adding new aspects.

Recent Issues