The rapid advancement and mass adoption of battery electric vehicles represent a pivotal shift in the global automotive industry, driven by the urgent need to address environmental sustainability and energy security concerns. As a core component of a battery electric vehicle’s powertrain, the performance, reliability, and efficiency of the drive motor directly influence the vehicle’s range, acceleration, safety, and overall user experience. Therefore, rigorous and accurate testing of these motors before they are integrated into a battery electric vehicle is not merely a quality control step but a critical necessity for ensuring the long-term viability and safety of electric mobility.
Traditional testing methods often rely on disparate equipment and manual data recording, which can be time-consuming, prone to human error, and lack the capability for sophisticated real-time analysis and predictive diagnostics. To overcome these limitations and meet the demanding production schedules of modern battery electric vehicle manufacturers, we have designed and implemented an advanced, automated test system. This system is built upon a synergistic integration of a Siemens S7-1200 PLC for robust hardware control and data acquisition, and National Instruments LabVIEW software for powerful graphical system design, data processing, visualization, and management. The primary objectives are to enhance testing precision, improve operational efficiency, enable comprehensive performance analysis, and incorporate intelligent predictive maintenance features for the drive motors used in battery electric vehicles.
The necessity for such a system stems from the complex operational profile of a battery electric vehicle. Unlike internal combustion engines, the drive motor in a battery electric vehicle must deliver instant torque, operate efficiently across a wide speed range, and withstand frequent start-stop cycles and regenerative braking events. Key performance parameters that must be validated include:
- Torque-Speed Characteristics: Mapping the motor’s output torque across its entire operational speed range.
- Efficiency: Determining the motor’s conversion efficiency from electrical input power to mechanical output power under various load conditions. The overall efficiency of the battery electric vehicle is heavily dependent on this.
- Thermal Performance: Monitoring and predicting temperature rise in the stator windings and permanent magnets to prevent insulation failure or demagnetization.
- Dynamic Response: Evaluating performance during transient conditions like acceleration and deceleration, which are frequent in battery electric vehicle driving.
Our system is designed to conduct a full suite of standardized tests, including no-load tests, load tests (both steady-state and dynamic), and simulated real-world driving cycle tests, all of which are essential for certifying a motor for use in a battery electric vehicle.
1. System Architecture and Hardware Design
The test system employs a modular and scalable architecture, logically divided into a supervisory control layer, a real-time control and acquisition layer, and the physical test bench layer. Communication between layers is facilitated via industrial Ethernet for high-speed, reliable data transfer within the test cell, and optional 4G connectivity for remote monitoring and diagnostics, which is particularly valuable for multi-site operations common in large battery electric vehicle supply chains.
The core hardware components are selected for their precision, reliability, and compatibility with the harsh electrical noise environment typical of power electronics and motor drives in a battery electric vehicle.
1.1 Programmable Logic Controller (PLC)
The central control unit is a Siemens SIMATIC S7-1200 (CPU 1214C). This PLC was chosen for its robust performance, integrated PROFINET interface, and sufficient I/O capacity. It executes the test sequences, interfaces directly with sensors and actuators, performs preliminary data processing, and serves as the communication gateway to the LabVIEW-based host computer. Its deterministic control loop is crucial for maintaining precise speed and torque setpoints during dynamic tests that simulate battery electric vehicle driving conditions.
1.2 Sensor Suite for Comprehensive Measurement
Accurate measurement is the foundation of any credible test system. For characterizing a battery electric vehicle drive motor, we employ a suite of specialized sensors:
| Parameter | Sensor Type / Model | Key Specifications | Purpose |
|---|---|---|---|
| Three-Phase Voltage & Current | Hall-Effect Sensor (e.g., ACS756) | Isolated measurement, high bandwidth, measures AC/DC waveforms. | Calculate electrical input power ($P_{in} = \sqrt{3} \times V_{LL} \times I_{L} \times \text{PF}$). |
| Torque & Speed | Reaction Torque Sensor / In-line Torque Flange (e.g., ZJ-A type) | Non-contact measurement (strain gauge or magnetic), high accuracy, low inertia. | Measure mechanical output power ($P_{out} = \tau \times \omega$, where $\omega = \frac{2\pi n}{60}$). |
| Temperature | PT100 Platinum Resistance Thermometer (RTD) | High accuracy and stability over a wide range (-100°C to 400°C). | Monitor winding and bearing temperature for thermal protection and loss analysis. |
| Vibration (Optional) | Accelerometer | High sensitivity, wide frequency range. | Diagnose mechanical issues like bearing wear or imbalance. |
The mechanical power is derived from the fundamental equation:
$$P_{out} = \tau \cdot \omega$$
where $\tau$ is the torque in Newton-meters (Nm) and $\omega$ is the angular velocity in radians per second (rad/s). The rotational speed $n$ in revolutions per minute (rpm) is related by $\omega = \frac{2 \pi n}{60}$.
The motor’s efficiency $\eta$, a paramount figure of merit for a battery electric vehicle, is then calculated in real-time:
$$\eta = \frac{P_{out}}{P_{in}} \times 100\%$$
This calculation is performed continuously throughout the test cycle.
1.3 Test Bench Actuators
The system features a dynamometer setup. The Device Under Test (DUT), the battery electric vehicle drive motor, is mechanically coupled to a load machine (another motor or an eddy current brake) controlled by a four-quadrant programmable drive. This allows the system to not only absorb power but also to simulate inertial loads and even drive the DUT, enabling tests for regenerative braking capability—a key feature of any battery electric vehicle.
2. Software Design and Implementation
The software architecture is designed for intuitiveness, flexibility, and powerful data handling. It is bifurcated into the lower-level PLC programming and the upper-level Human-Machine Interface (HMI) and data management system.
2.1 PLC Programming with TIA Portal
The control logic for the S7-1200 is developed using Siemens TIA Portal (V15.1). The program is structured into well-defined function blocks (FBs) and organization blocks (OBs):
- Main Cyclic OB (OB1): Executes the core state machine that controls the test sequence (Idle, Initiate, Ramp-up, Steady-State Measurement, Transient Test, Ramp-down, Safe Stop).
- Communication FB: Manages the data exchange protocol with the LabVIEW host computer over TCP/IP.
- PID Control FB: Implements closed-loop control algorithms for precise regulation of motor speed and torque, critical for replicating the dynamic loads experienced by a battery electric vehicle.
- Safety OB (OB121): Handles fault conditions (overcurrent, overtemperature, communication loss) to ensure a safe shutdown.
2.2 Host Computer System with LabVIEW
LabVIEW serves as the central hub for user interaction, data visualization, analysis, and storage. Its graphical programming paradigm (G-code) is ideal for creating custom test and measurement applications.
a) Test Sequence Management: The front panel provides a hierarchical menu for selecting test types relevant to battery electric vehicle motor validation:
- No-Load Test (to determine iron losses and friction losses).
- Load Test (steady-state efficiency mapping).
- Dynamic Cycle Test (simulating standard driving cycles like WLTP or FTP-75).
- Overload and Short-Duration Tests.
- MTPA (Maximum Torque Per Ampere) characteristic verification.
b) Real-Time Data Visualization: The interface displays real-time trends of all critical parameters (Torque, Speed, Voltage, Current, Power, Efficiency, Temperature) using charts and graphs. Digital indicators show instantaneous values. This allows the operator to monitor the health of the battery electric vehicle drive motor during the entire test procedure.
c) Data Logging and Database Integration: All acquired data is time-stamped and streamed to a relational database. We utilize Microsoft Access via ODBC connectivity for this purpose. LabVIEW’s database toolkit enables seamless execution of SQL commands to create tables, insert records, and perform queries. This structured storage allows for efficient post-test analysis, report generation, and long-term traceability of every battery electric vehicle motor tested. A snapshot of the data management module is conceptually shown below:
[A LabVIEW block diagram segment would show a while loop containing: DAQ Read -> Data Scaling -> Bundle into Cluster -> Build SQL INSERT String -> DB Tools Open Connection -> DB Tools Execute Query -> DB Tools Close Connection].
3. Intelligent Temperature Prediction Model Using BP Neural Network
Thermal management is a critical challenge in battery electric vehicle motors. Excessive temperature rise leads to permanent magnet demagnetization, insulation degradation, and ultimately, motor failure. While direct measurement with RTDs is reliable, it is reactive. We have implemented a proactive predictive model using a Backpropagation (BP) Neural Network to forecast the motor’s winding temperature based on its operating state.
The BP neural network is a multilayer feedforward network trained by error backpropagation. It excels at modeling complex, non-linear relationships—ideal for mapping motor operational parameters to thermal response.
3.1 Model Architecture and Algorithm
We designed a network with one input layer, two hidden layers, and one output layer.
- Input Layer (4 neurons): The selected features that most influence temperature are:
- Motor Torque ($\tau$)
- Motor Speed ($n$)
- Current Motor Winding Temperature ($T_{current}$)
- Ambient Temperature ($T_{ambient}$)
- Hidden Layers: Two layers with 10 and 5 neurons respectively, using a sigmoid activation function $f(x) = \frac{1}{1 + e^{-x}}$ to introduce non-linearity.
- Output Layer (1 neuron): The predicted future winding temperature ($T_{predicted}$) after a defined time interval (e.g., 60 seconds), using a linear activation function.
The forward propagation for a single neuron is:
$$y_j = f\left(\sum_{i=1}^{n} (w_{ji} \cdot x_i) + b_j\right)$$
where $y_j$ is the neuron’s output, $x_i$ are its inputs, $w_{ji}$ are the connection weights, $b_j$ is the bias, and $f$ is the activation function.
The core of the training lies in backpropagation. The error $E$ between the network’s prediction $T_{pred}$ and the actual measured temperature $T_{actual}$ is calculated using Mean Squared Error (MSE):
$$E = \frac{1}{N} \sum_{k=1}^{N} (T_{actual}^{(k)} – T_{pred}^{(k)})^2$$
This error is then propagated backward through the network, and the weights ($w_{ji}$) and biases ($b_j$) are adjusted using gradient descent optimization to minimize $E$. The weight update rule is:
$$w_{ji}^{(new)} = w_{ji}^{(old)} – \alpha \cdot \frac{\partial E}{\partial w_{ji}}$$
where $\alpha$ is the learning rate.
3.2 Model Training and Validation
Historical test data from numerous battery electric vehicle drive motor trials was used to train the network. The dataset was normalized and split into training (70%), validation (15%), and testing (15%) sets. The training process continued until the error on the validation set stopped improving, preventing overfitting.
The performance of the model was excellent. A comparison between the predicted temperature and the actual measured temperature from an unseen test dataset shows a close correlation, as conceptually illustrated in the plot below. The predicted curve effectively tracks the actual thermal dynamics of the battery electric vehicle motor.
The strength of the linear relationship between predictions and actual values is quantified by the coefficient of determination, $R^2$. A value close to 1 indicates a highly accurate model. Our model achieved an $R^2$ value of:
$$R^2 = 0.93029$$
This high value confirms the model’s reliability for temperature prediction in a battery electric vehicle drive motor, enabling early warnings before critical temperature thresholds are reached.
4. System Testing and Experimental Results
The complete test system was validated through a series of exhaustive experiments on a 50 kW permanent magnet synchronous motor (PMSM) typical of those used in mid-size battery electric vehicles.
4.1 Steady-State Load Test and Speed Control Accuracy
The motor was operated at its rated condition (2935 rpm, 35.8 Nm). The system successfully logged all parameters and displayed them in real-time. A key test of the control system’s precision is its ability to accurately hold and step through speed setpoints. We commanded the load machine to control the DUT’s speed from 0 to 2800 rpm in increments of 400 rpm. The commanded vs. measured speed data is tabulated below:
| Commanded Speed (rpm) | Measured Speed (rpm) | Absolute Error (rpm) |
|---|---|---|
| 0 | 0.1 | 0.1 |
| 400 | 399.7 | -0.3 |
| 800 | 800.4 | 0.4 |
| 1200 | 1199.5 | -0.5 |
| 1600 | 1600.8 | 0.8 |
| 2000 | 1999.9 | -0.1 |
| 2400 | 2400.2 | 0.2 |
| 2800 | 2801.1 | 1.1 |
The results demonstrate exceptional speed control accuracy, with all errors well within ±2 rpm. This far exceeds the control precision requirement specified in standards such as GB/T 18488-2024 (e.g., ±50 rpm for speeds below 10,000 rpm), confirming the system’s suitability for high-precision testing of battery electric vehicle drive motors.
4.2 Efficiency Mapping
A crucial output for any battery electric vehicle motor test is its efficiency map. The system automatically conducted a grid of tests across the motor’s torque-speed envelope. The resulting efficiency map, contour-plotted by post-processing the logged database, clearly shows the high-efficiency region—a vital input for battery electric vehicle powertrain optimization to maximize range.
4.3 Neural Network Temperature Prediction Validation
During a prolonged load cycle test, the BP neural network model was run in parallel with direct RTD measurements. The model successfully predicted temperature trends, providing advance warning of a thermal rise approximately 45-60 seconds before it was fully reflected in the direct sensor reading. This demonstrates the practical value of integrating AI-based predictive analytics into the test system for battery electric vehicle components.
5. Conclusion and Future Work
We have successfully designed and implemented a high-precision, automated, and intelligent test system for battery electric vehicle drive motors. The system integrates robust PLC-based control with the flexible and powerful data handling capabilities of LabVIEW. It fulfills all standard testing requirements—from basic no-load tests to complex dynamic driving cycles—with superior accuracy and efficiency. The incorporation of a BP neural network for temperature prediction adds a valuable layer of diagnostic intelligence, moving beyond simple monitoring towards predictive health management for the battery electric vehicle motor.
The system offers a significant advancement over conventional test stands by providing a unified platform for control, measurement, analysis, and data management. Its design is directly responsive to the growing quality and throughput demands of the global battery electric vehicle industry. Future work will focus on integrating more advanced AI models for fault diagnosis (e.g., detecting eccentricity or winding shorts), expanding the system to test the complete electric drive unit (EDU) including the inverter, and enhancing the cloud connectivity for fleet-wide performance analytics of battery electric vehicle drive systems.