This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TMS320F28P550SJ: I'm not quite sure about some of the data about Motor Fault Detection.

Part Number: TMS320F28P550SJ
Other Parts Discussed in Thread: C2000WARE

Tool/software:

Hi ALL

(1)I've found the data in the sample project. What do vibx, viby, and vibz represent respectively?

(2)I don’t fully understand these preprocessing parameter settings. If you could explain them to me, I would really appreciate it.

Best Wishes,

Xia

  • Hi Xia,

    Please see my comments below:

    (1) Vibx, Viby, and Vibz represent the value of vibration sensor in x, y, z axis. 

    (2) The preprocessing preset determines what preprocessing method will be applied to the data before training a model. You probably have noticed other parameters change accordingly when you select different items from the drop-down menu. You can also change the parameters in other fields, in this case, the preprocessing method changes automatically to "5. Custom_MotorFault". For the supported preprocessing methods and an explanation to the parameters, please refer to the user's guide.

    Thanks,

    Jiaxin

  • Hi Jiaxin,

    Thank for your reply!I have some questions I'd like to ask you:

    (1)What do the '10Hz' and '0Nm' in the red box represent? What does '10Hz' specifically mean in terms of frequency?

    (2)What is the difference between 1D and 2D convolutions? How do they affect training results?

    The user manual only mentions that the 1D channel is 1, and the 2D1 channel is 3. But haven’t we already set the Channel to 3? What is the difference?

    Thanks,

    Xia

  • (1) The "Hz" and "Nm" are the operation condition of the motor. "Hz" is the motor's electric frequency

    The user manual only mentions that the 1D channel is 1, and the 2D1 channel is 3. But haven’t we already set the Channel to 3? What is the difference?

    (2) This field determines how the preprocessed features are organized and fed into the AI model. Yes, we have 3 axis of data, but you can extend them into a 1-dim vector n or stack them to a 3-dim vector.

    Thanks,

    Jiaxin

  • Hi Jiaxin,

    Nice to receive your reply‌. I have the following questions hoping for your response:

    (1)I currently have a TMS320F28P55 development board but lack a vibration sensor. I've trained and compiled the example motor fault detection model. Now I want to load the model onto the board—how can I achieve this?

    (2)How is the model stored?

    (3)In which memory space is it stored?

    (4)How do the main CPU and NPU collaborate during operation?"

    Thank you again for your patient reply.

    Thanks,

    Xia

  • Hi Xia,

    (1) For model deployment, please refer to Chapter 4.3 of the user's guide. Did you download the c2000ware_motor_control_sdk? There are software examples to help you understand the whole flow and assist with your design.

    (2) The compiled model is the ".a" file in the artifacts folder

    (3) Please check the .cmd file in the example projects to know where the model is stored.

    (4) NPU is an independent hardware accelerator. The model operation is invoked by calling the function "tvmgen_default_run()" and the model will be executed independently to the CPU then. When it is finished, a flag will be set to inform CPU about the completion. You can also confirm the logic with the example code.

    Thanks,

    Jiaxin

  • Hi Jiaxin,

    Do I currently not have a vibration sensor to perform this step?

    Thanks,

    Xia

     

     

  • No, you don't need a vibration sensor to duplicate this step. What the example project does is to have the test vector computed once in the main() and pass the result for the live preview in an emulated 4kHz interrupt.

    Thanks,

    Jiaxin