Engineers say | The future of digital motor control: multiple motors on one MCU, embedded AI, and advanced algorithms

Figure 1 MCK-RA8T1 with two inverter boards

For controlling two independent motor assemblies

This setup demonstrates the detection of two anomalies that occur in real motor operation: shaft misalignment and load unbalance conditions. Shaft misalignment occurs when the motor’s output shaft is not perfectly aligned in space with the load it is driving, resulting in unintended torque and friction. Load unbalance conditions occur when the system driving the load is out of dynamic balance, generating oscillatory forces as the motor rotates orthogonally to the shaft, and oscillations in the form of load resistance changes during each revolution.

Both of these conditions are serious issues for motor systems and can lead to a range of serious problems, from power loss to excessive noise and vibration to rapid bearing wear and dangerous bearing and structural failures. In some applications, such as high-power industrial systems, even small deviations from proper alignment and balance over time can cause significant concerns. In other applications, such as washer dryers, a certain amount of imbalance and misalignment is expected, but exceeding it can lead to internal machine collisions and failures.

To create shaft alignment issues in a small area, our first motor was configured with a shaft that connected it to the gearbox. When the circuit board holding the motor was slightly deformed by a push from a finger, the shaft bent and moved out of alignment with the gearbox.

To demonstrate the balance issue, our second motor is configured with an aluminum hub that we have the option of inserting small M4 screws into. Without the screws, the hub provides a normal, balanced load to the drive motor. With the screws added, the motor load is unbalanced by a small amount.

Figure 2 Status data sent from the Board dashboard

Our two AI modules have been trained to detect both of these situations. We then added the communications to the dashboard display (as shown in Figure 2) so that we can see what is happening in real time.

The display marks the condition of each motor, with status indicated by color codes and text. It also includes a scrolling timeline that lights up orange whenever the system sees an anomaly about a certain threshold (i.e., a misaligned or unbalanced condition).

As shown in Figure 3, there are quite a few processes running simultaneously on the RA8 MCU. We have two independent instances of the vector motor control algorithm driving the BLDC motor. Each of these instances also sends real-time data from the control process to two coupled AI inference modules. The AI ​​modules monitor this data for indications of normal versus problematic behavior. Since only internal motor control data is used, the AI ​​system does not require additional sensors to monitor for anomalies. This is a powerful feature of RealityCheck ^™ Motor Toolbox , allowing engineers to create AI monitoring solutions without adding any sensor BOM cost to their designs.

As shown in Table 1, our AI modules have extremely small RAM and flash memory footprints, and they are both very fast, leveraging the advanced features of the RA8T1. The total flash memory used for both is less than 14KB, and the RAM is less than 5KB. Each module performs an inference in 1 millisecond, so even with 8 inferences per second for each motor subsystem, the load on the CPU is very small.

Figure 3 Process and data flow in two motor AI examples

Table 1 Embedded model indicators

The story begins with Renesas providing e² studio users with a seamless, integrated workflow for interacting with Reality AI tools , all as part of our RealityCheck ^™ Motor Toolbox. The functionality is shown in Figure 4 below. Engineers can connect to a Renesas motor control MCU and collect data directly into our Data Storage Tool plugin. This can be done using an evaluation board such as the MCK-RA8T1 or in their own Renesas MCU-based hardware design. Our toolbox provides example code tied to the Renesas Flexible Software Package (FSP) stack, guiding users through selecting and capturing real-time data parameters from vector motor control algorithms. In minutes, users can collect real-time data from their boards under actual motor usage conditions.

Specifically for unbalanced load detection, our data collection matrix has two columns of variation: (1) the presence and absence of a screw in the hub, and (2) the range of speeds we want to address. To build our demonstration, we collected data samples for combinations of these conditions, and repeated them multiple times to cover the random variations that occur in any physical setup. Table 2 shows an example data coverage matrix.

Figure 4 RealityCheck Motor toolbox for creating motor AI

Provides a seamless round-trip workflow

Table 2 Data collection plan for unbalanced load detection

Figure 5 e-AI Translator allows you to easily import AI models from open source platforms such as Keras, Tensor Flow, and PyTorch as code in your e² studio project

Some customers prefer to bring in their own AI models, developed in well-known open source frameworks such as Keras, Tensor Flow or Pytorch. RealityCheck Motor lets you do this too with the help of the Renesas e-AI Translator plugin. This powerful e² studio plugin enables users to import open source created AI, converting them into compact C code that can be directly linked into your project.

This is how we built our shaft misalignment detector. We once again collected data in a similar way as before, using the RealityCheck Motor data acquisition and data storage tools in e² studio. However, instead of going through Reality AI Tools, we imported this new data into the Python framework and trained our demo AI model there using Keras. We then imported the fully trained Keras model into our project code using the e-AI Translator. The model was already configured to handle the data stream from our RealityCheck Motor capture function, making it easy to integrate the e-AI inference code with the rest of the project.

This example shows how we can easily combine AI modules created by the Reality AI Tools workflow with AI modules imported from other sources via the e-AI converter. Users can import any deep learning or NN model from common training frameworks, so if you prefer to bring your own model, we are happy to help.

Renesas Electronics (TSE: 6723)