“CREATOR case”: PMSM and IM electric machine data for validation and benchmarking of simulation and modeling approaches Open Access

The tedious design process of electric machines that involves different disciplines, e.g. electrical, mechanical, magnetic and thermal aspects, and also increasingly dynamic analyses, calls, for example, cases for benchmarking of modeling and simulation approaches. Today, the literature only presents selected machine design-related data and shows the laboratory results of chosen parameters or performance criteria. As verification and validation of models has become increasingly important, so has the need for the availability of data to allow for verification and validation of models, as well as reproducibility and reusability of results (Oberkampf and Roy, 2010). To this aim, and to the authors’ best knowledge for the first time, we publish the complete sets of machine data for two different machines for such multifaceted analysis. These packages of data provide all the required information in terms of design parameters and measurement results to facilitate their use as validation and benchmarking tool for different modeling approaches, beyond the specific quantitative results as obtained from singular analyses. For example, Bergfried et al. (2023) have investigated the specific question of thermal modeling and simulation based on our data.

Electric machines, as key players in energy conversion, have been known for more than a century. The recent technological advancements in electric drives and power electronic systems, together with new materials and manufacturing techniques and increasingly transient performance requirements, provide plenty of opportunities for future innovations, but also increase the demands on the modeling and simulation tools, e.g. Ahn et al. (2023) and Hwang et al. (2021). Current electric machine design procedures usually start with an expert’s choice for a particular machine type and topology or already-developed design choices, e.g. Barcaro and Bianchi (2013), Krause et al. (2013) and Mese et al. (2016). Optimization is then typically based on selected parameters and steady-state operating points, e.g. Rimpas et al. (2023), Gobbi et al. (2024) and Li et al. (2017). However, exploiting the full potential of electric machines in the future requires more powerful approaches to account for the different design possibilities, parameters and criteria. We provide all data relevant for the comprehensive modeling of two machines, as well as measurement results for both selected steady-state operating points and three different drive cycles, to allow for code validation or benchmarking, e.g. as a baseline for innovative shape or topology optimization techniques. As a matter of fact, well-prepared experimental data is incredibly valuable for the validation of simulation models and available data can underpin the reliability and reproducibility of the results for the community (Weinper and Tomczak, 2021). We thereby also contribute to increasing availability of results in the field of electric machines that follow the FAIR (Findability, Accessibility, Interoperability and Reusability) principles, which enables transparency and reproducibility of research knowledge, an acronym introduced almost a decade ago by a consortium of scientists and organizations (Wilkinson et al., 2016). In contrast to the reporting of selected results without context, e.g. on data science competition platforms (Fedesoriano, 2021; Shimizu, 2023; Ferretti, 2022) and to the problems formulated by the TEAM (Testing Electromagnetic Analysis Methods) initiative (TEAM Problems – International Compumag Society, 2018) or the Galileo Ferraris Contest (Gallileo Ferraris Contest, 2024), we are not posing a specific challenge.

The Collaborative Research Centre “Computational Electric Machine Laboratory - CREATOR” (CRC – TRR361/F90) (Österreichischer Wissenschaftsfonds FWF, 2022; Deutsche Forschungsgemeinschaft DFG, 2022) aims at enhancing electric machine performance and computational efficiency by bringing together advances in many different disciplines. As part of this project, essential design and measurement data of electric machines are obtained and shall be made available to many other researchers within the community. The work that generates the data on these example motors itself assesses the performances of different modeling approaches to analyze dynamic drive cycle operation, including the crucial experimental validation.

The two motors are available and analyzed in the Electric Drives and Power Electronic Systems Institute (EALS) laboratory at TU GRAZ, an induction motor (IM) and a permanent magnet synchronous motor (PMSM). While the machines were originally not optimized for traction applications, both are machines designed for research purposes, with all relevant data available and some characteristics that make them notably interesting for research – and also benchmarking – purposes.

The machine design and measurement data packages are made available through two central repositories, Heidarikani (2024) and Dhakal (2024), one for each of the two machines. They are discussed and analyzed in Sections 5 and 6, respectively, for each of the two machines. Prior to this, the next section outlines the organization of these two main parts of the paper.

Subsections 5.1 and 6.1 detail the repositories of the design data parameters for both the IM and the PMSM, respectively. They provide a structured framework for analysis and comparison. Within these subsections, each part is dedicated to a specific aspect of the motors’ designs, as follows: Subsections 5.1.1 and 6.1.1 detail the motor geometries, Subsections 5.1.2 and 6.1.2 the material properties, Subsections 5.1.3 and 6.1.3 the electrical parameters and Subsections 5.1.4 and 6.1.4 the winding schemes of the IM and the PMSM, respectively.

In Subsections 5.2 and 6.2, the repository also provides comprehensive measurement results concerning both the IM’s and the PMSM’s conventional operating characteristics such as no-load tests for both motors (Subsections 6.2.1 and 5.2.1, respectively) and the locked rotor test specifically for the IM (Subsection 5.2.1), as well as the derivation of equivalent circuit parameters from measurement data (Subsections 5.2.2 and 6.2.2, respectively).

In addition to these steady-state characteristics, the repository provides extensive measurement results of drive cycles, i.e. predefined sequences of torque-speed pairs over time for the analysis of dynamic drive performance. Three standard reference drive cycles are selected to evaluate the performance of electric machines in traction electrification: the WLTP (worldwide harmonized light vehicles test procedure), the Braunschweig city drive cycle (urban) and the Artemis 130 km/h drive cycle (inter-city) (DieselNet, 2011). The WLTP offers a balanced representation of urban, suburban and highway driving scenarios. The urban drive cycle emphasizes low-speed, high-torque conditions, crucial for assessing city driving performance characterized by frequent stops and starts. The inter-city drive cycle examines moderate speeds and less frequent stops, reflecting typical conditions encountered during inter-city commutes.

To derive the required torque and speed for each drive cycle, two reference vehicles are considered: the BMW i3 (medium range) and the Smart EQ (small range), employing their specifications in quasi-static longitudinal vehicle models (Guzzella and Amstutz, 2005). Inputting the drive cycle into the longitudinal vehicle model of each EV determines the required torque and speed on the wheels of the vehicle. The necessary torque and speed of the vehicle motor can then be determined by utilizing the gear transmission. However, since the rating of the vehicle motor may differ from that of the motors in the laboratory, a down-scaling method is employed to ensure that the input data falls within the range of the IM and the PMSM in the laboratory. Detailed information about the down-scaling method, the longitudinal vehicle model and vehicle data is presented in Dhakal et al. (2023).

Schematic overviews of the test benches and the control architectures for both motors are provided in Appendices 1 and 2, respectively.

Subsections 5.2.3 and 6.2.3 discuss the input drive cycle torque and speed data for the six cases (three cycles, two vehicles) per motor as well as the corresponding measurement results for both machines, including the output and input power and the measured losses of the IM and of the PMSM, respectively.

The three-phase squirrel cage IM is completely enclosed and water-cooled to maintain optimal performance under varying conditions. With a total of 92 temperature sensors strategically integrated, this motor enables comprehensive temperature analysis, as it was originally manufactured for analysis of temperature distribution under post-fault operation (Eickhoff et al., 2021).

All IM related data and experimental results are made available online through the central repository Heidarikani (2024). Table 1 summarizes the general parameters of the IM.

Figure 1 shows a 2D CAD drawing of the motor, with complete dimensions, detailing the motor housing, rotor, stator and windings. Additional geometric data essential for modeling the IM, such as details on the end ring, end winding and housing, are provided in  Appendix 1.

The online repository also contains pre-built geometry model files for ease of use.

Table 2 shows the materials used in the different components of the IM. For each material, the corresponding online data repository link is cited.

Table 3 provides the electrical parameters of the IM.

An overview of the winding scheme is presented here. To this aim, Table 4 details the parameters of the winding of the IM. A figure visually illustrating the coil designations for all phases, showing the winding arrangement within the motor is provided in the repository.

The no-load test evaluates the IM performance without mechanical load. From this, selected parameters such as core loss, magnetization current and magnetization inductance can be identified. Conversely, in the locked-rotor test, the motor is mechanically prevented from rotating, and voltage, respectively currents are applied to determine impedance parameters such as rotor resistance and leakage inductances. For more explanations on these tests, see, e.g. Hendershot and Miller (2010) and Pyrhonen et al. (2009).

Both tests were carried out for different supply frequencies and amplitudes.

To separate the iron and the sum of friction and windage losses, the measurement results were separated into their frequency-dependent and frequency-independent parts. The results are also available in the repository.

The equivalent circuit parameters of the IM, as derived from the measurements, are shown in Table 5. The stator resistance (R_s) was measured through a direct current test, the rotor resistance (R_r) determined from the locked-rotor test. The stator magnetization inductance (L_m) was derived from a no-load test, and the combined leakage inductance (L_σs + L_σr) was calculated from the locked-rotor test. Due to motor symmetry, the two leakage inductances L_σs and L_σr are assumed to be equal (Tessarolo et al., 2015).

Figure 2 shows the measured magnetization flux (λ_m) versus current (I_m) and the derived current-dependent inductance (L_m). Figure 3 displays the variation of the rotor resistance with frequency, as well as the relationship between stator and rotor leakage inductances and frequency. The low-frequency single-phase equivalent circuit of the IM is shown in  Appendix 1.

An exemplary case of the analysis of the WLTP class 3 drive cycle for the IM, using the down-scaling procedure (Dhakal et al., 2023) based on the specifications of a medium-sized vehicle (BMW i3), is presented. It covers both the input data for the measurement, i.e. the torque and the speed, as well as the measured output data of the IM throughout the entire drive cycle. The measured input and output powers, along with the total loss calculated from the measurement, as well as the input torque and speed profiles are shown in Figure 4. The online repository is not limited to the single drive cycle case depicted in this example; it holds complete measurement data for the six cases of both vehicles and three drive cycles.

The PMSM machine presented here is a prototype inset PM machine originally designed for use in high-efficiency household cooling appliances with a maximum continuous power of 70 W and a targeted optimum performance at 7 W. With a maximum continuous power and maximum efficiency occurring at different operating points, it reflects demands on many modern motors, notably used in traction applications.

The prototype machine is naturally cooled and designed to run long hours with considerably high efficiency. All PMSM related data and experimental results are made available online through the central repository Dhakal (2024). Table 6 describes the general design parameters of the PMSM.

A 2D geometry diagram of the PMSM is illustrated in Figure 5. The stator, rotor and magnet dimensions are highlighted. Table 12 details the motor geometry. For ease of use, pre-built geometry model files are also available in the online repository.

Table 7 details the motor parts and the corresponding material used. For each material, the corresponding online repository link is cited.

Table 8 provides the electrical parameters of the PMSM.

Here, the winding scheme of the PMSM is presented. Table 9 details the winding properties. A figure showing its model and the slot view is provided in the repository.

No-load tests of the PMSM are carried out to assess the load-independent losses within the motor itself, such as iron and frictional losses, and to assess the machine parameters, such as back-EMF and cogging torque. For more explanations on such kinds of tests, see, e.g. Hendershot and Miller (2010) and Pyrhonen et al. (2009). The PMSM is tested under no-load, both with the rotor mounted within the stator and without rotor (to identify the iron loss), at different rotor speeds. Figure 6 shows the no-load torque measurement results as function of speed and the calculated no-load iron losses as function of supply frequency.

For the measurement of the cogging torque, the rotor is rotated very slowly, at 0.25 rpm. The torque measurements are taken by a torque transducer. To measure the back-EMF, the rotor was rotated at a constant speed of 2000 rpm. The results are also provided in the repository. A fundamental peak component of the induced back-EMF of 47.5 V per phase is identified by discrete Fourier transform.

Table 10 shows the low-frequency equivalent circuit parameters of the PMSM. The stator resistance is measured directly using an LCR meter (LCR meter, 2023). The d- and the q-axis inductances are obtained through finite element analysis (FEA) using the JMAG^® software’s (JSOL Corporation, 1983) inductance calculator tool (JSOL Corporation, 2022). An exemplary 2D JMAG model file used for this analysis is also made available in the repository.

To calculate the current angle for maximum torque, the current angle was varied from 90°el to 120°el, with current amplitudes of 0.15 and 0.3 A. The motor was set to operate under motoring operating points at a constant speed of 2000 rpm. The torque was measured as the current angle was varied and the angle that gives the maximum torque value is identified. The maximum current angles with current amplitude of 0.15 and 0.3 A are found to be 100°el and 110°el, respectively. The low-frequency single-phase equivalent circuit models of the PMSM are provided in  Appendix 2.

Exemplarily, the measurement results of the WLTP class 3 drive cycle with reference to the medium-sized passenger car (BMW i3), down-scaled to the PMSM as per Dhakal et al. (2023), are shown here, see Figure 7. They include the drive cycle input speed and torque, the input power, the output power and the total losses throughout the whole drive cycle. The online repository contains the complete set of measurement data for both vehicles and all three drive cycles.

This paper presents comprehensive electric machine design parameters and measurement results of two different electric machines and for a large set of six drive cycles per machine. It makes all relevant design data available for benchmarking of modeling and simulation approaches. This paper not only describes the data itself but also guides the prospective user through their organization within the repository.

This work was partially supported by the joint Collaborative Research Centre CREATOR (DFG: Project-ID 492661287/TRR 361; FWF: 10.55776/F90) at TU Darmstadt, TU Graz and JKU Linz. The authors would like to thank Dr Hermann Schranzhofer from TU Graz for his help with the realization of the repositories as well as helpful suggestions on the overall organization of the open science contribution of these data. They would also like to thank Mario Mally and Michael Wiesheu from TU Darmstadt for through proofreading of an earlier version of this paper.

The detailed geometric parameters of the IM are listed in Table A1.

Equivalent circuit model of the IM

As is common for simplicity, the IEEE single-phase equivalent circuit model (ECM) of the IM is presented without considering core losses, as shown in Figure A1 (Kojooyan-Jafari et al., 2015). The diagram illustrates the per-phase equivalent circuit of the IM, where the rotor-side impedances are converted to the stator side. In the circuit, R_s and R_r represent the stator and rotor resistances (referred to the stator side), while L_σs and $Lσr′$ denote the stator and rotor leakage inductances. The magnetization inductance is represented by L_m. In addition, the stator voltage V_s, rotor voltage V_r, stator current I_s and rotor current I_r are depicted. The values of the equivalent circuit parameters for the IM can be found in Table 5.

Laboratory test setup and control architecture of the IM.

Rotor field-oriented control (RFOC) is used to control the speed of the IM, here the device under test (DUT). In addition, the torque control is applied to the PMSM, which functions as the load machine. The control architecture illustrated in Figure A2 outlines the comprehensive control scheme used in the laboratory’s IM test bench. Both inverters use space vector pulse width modulation (SVPWM), at a constant switching frequency of 5 kHz.

Table A2 lists the detailed geometry parameters of the PMSM.

Equivalent circuit model of the PMSM.

In the majority of analyzed PMSM cases, single-phase ECMs are presented, disregarding the core loss component (Ba et al., 2022). Simplified ECMs of a PMSM are demonstrated in Figure A3. Figure A3 (a) presents the single-phase ECM of a PMSM. As presented in the diagram, R_s is the per-phase resistance of the stator winding and L_s is the synchronous inductance, which is an equivalent inductance of self and mutual per-phase inductances. The flux linkage of permanent magnets is denoted by λ_pm. The back electromotive force, E₀ is proportional to the electrical rotational frequency ω_e. The single-phase current and voltage are denoted by I_p and V_p, respectively. Figure A3 (b) and (c), represents the corresponding d and q-axis ECMs, respectively. As denoted in the diagram, V_d and V_q are the d- and q-axis terminal voltages, and I_d and I_q are the d- and q-axis armature currents. The d- and q-axis inductances are denoted by L_d and L_q, respectively. The corresponding equivalent circuit parameter values of the PMSM can be refereed from Table 10.

Laboratory test setup and control architecture of the PMSM.

To control the drive cycle’s input speed and torque, a cascaded control technique is used. Figure A4 illustrates the overall control scheme of the PMSM test bench in the laboratory. The DUT, which is the PMSM presented here, is torque-controlled, and the load machine, which is also a PMSM, is speed-controlled. The choice to torque-control the DUT was based on the goal of obtaining fewer harmonics in the input currents at the machine terminal. Both machines side inverters operate at a switching frequency of 20 kHz. A maximum torque per ampere (MTPA) control scheme is used for the current controllers.