NXP Application Code Hub

Time Series Studio powered, on-device learning Fan anomaly detection based on FRDM-MCXA156

An application designed for monitoring the status of the fan, which utilizes an accelerometer sensor attached to the fan to check if the fan is operating in a normal condition. Application is based on FRDM-MCXA156 with 96MHze Cortext-M33, 1M Flash and 128KB SRAM inside

The Appliccation is powered by IKM(Incremental KMeans for anomaly detection) model, which is trained and generated by eIQ Time Series Studio and support on-device learning as well. Model size is 7 KB and RAM requirement is 4 KB Model inference during is 6ms

How to train the model, please refer to AN14549 from nxp.com.

Below charts are the block of system and software workflow.

Boards: FRDM-MCXA156

Categories: AI/ML, Anomaly Detection

Peripherals: DISPLAY, I2C

Toolchains: MCUXpresso IDE

Table of Contents

1. Software
2. Hardware
3. Setup
4. Results
5. FAQs
6. Support
7. Release Notes

1. Software

• Download SDK_2_16_0_FRDM-MCXA156
• Download and install MCUXpresso IDE V11.9.0 or later.
• MCUXpresso for Visual Studio Code: This example supports MCUXpresso for Visual Studio Code, for more information about how to use Visual Studio Code please refer here.
• Download and install eIQ Toolkit 1.13.1

2. Hardware

• FRDM-MCXA156
• PAR-LCD-S035
• ACCEL-4-CLICK
• FAN
• Type-C USB cable

3. Setup

3.1 Step 1

• Connect board with LCD

  

• Mounting ACCEL-4-CLICK to the FAN

  

• Connect board with ACCEL-4-CLICK

  

• Connect debug port

  

• whole system

  

3.2 Step 2

Import "dm-tss-powered-on-device-learning-fan-anomaly-based-on-mcxa156" from Application Code Hub

• Import project in MCUXpresso IDE

  

• Import project in VS code

  

• Build project and download to the board

• Notice: please erase the whole flash before download the firmware

4. Results

• Turn on the Fan, then normal state is detected,Switch the speed of Fan, the state of Fan remain normal

  

• Knock on the Fan than anomaly is detected

  

• Disrupting the operation of the Fan blades, anomaly is detected

  

• Modify the mounting angle of ACCEL-4-CLICK by rotating it 90 degrees in the counterclockwise direction. An anomaly has been detected.

  

Training on Device

• Enter Trainer window, click start button ensure that don’t move the Fan during the training procedure

  

• Return to main after training is complete

  

• If the model fails to accurately recognize other speeds, then conduct the training once more.

5. FAQs

Insufficient on-device training epoches will lead to a significant decline in model accuracy. Solution: Increase the number of training epoches to achieve better accuracy.

If want to use the default model, click the button "Default" in the Trainer

6. Support

Project Metadata

Questions regarding the content/correctness of this example can be entered as Issues within this GitHub repository.

Warning: For more general technical questions regarding NXP Microcontrollers and the difference in expected functionality, enter your questions on the NXP Community Forum

7. Release Notes

Version	Description / Update	Date
1.0	Initial release on Application Code Hub	November 27th 2024

Licensing

LA_OPT_NXP_Software_License v58

Origin

if applicable - note components your application uses regarding to license terms - with authors / licenses / links to licenses, otherwise remove this section