Monitor refrigerator temperature using ESP32-DevKitC based on ML model

 Edge Impulse is the leading embedded machine learning (ML) development platform, supporting more than 1,000 global enterprises and 10,000 machine learning projects. Edge Impulse provides an easy-to-use end-to-end platform that enables:
- Easily input sensor data into embedded machine learning models;
- Efficiently create embedded machine learning models optimized for memory size and power consumption;
- Easily deploy models on target sensors using the Edge Impulse SDK and a development environment of your choice.
In a recent blog, software application engineer Zin Thein Kyaw introduced how to use Edge Impulse for embedded machine learning development, taking cold chain monitoring applications as an example. He built a refrigerator temperature monitor using an anomaly detection model based on the ESP32-DevKitC development board, the MAX31855 board, and a K-type thermocouple probe.
Temperature monitoring of refrigerators is important to ensure that perishable foods are stored refrigerated at the right temperature. Zin Thein Kyaw used the following components in this project for a proof-of-concept (PoC) demonstration:
-ESP32-DevKitC development board
-Adafruit MAX31855 thermocouple board
-K-type industrial thermocouple probe
-Embedded firmware compiled using the Arduino IDE, including Edge Impulse SDK
users can easily migrate the configuration of this PoC demonstration to other embedded devices. They can also view the data set of the Edge Impulse project and the training method of the anomaly detection model on Edge Impuose Studio.
A few minutes after opening or closing the refrigerator door, Edge Impulse creates an anomaly detection embedded machine learning model based on the temperature change curve of the sensor to detect the opening and closing status of the refrigerator door. At the same time, it can also obtain the duration of this state through the "abnormal score".
During the training of the anomaly detection model, you can use the "data forwarder" tool to collect data. The "data forwarder" is part of the Edge Impulse command line interface (or CLI) that establishes a secure network connection between the host PC, sensors, and the Edge Impulse platform, collects real-time sensor data, and uses it to train embedded machine learning Model.
Visualization of critical temperature ranges in "temperature anomaly areas" is implemented in Edge Impulse Studio. Used in Zin Thein Kyaw's project, it can detect outliers in the data and mark them as "anomalies", providing an important reference for embedded machine learning models using the K-means clustering algorithm.
In summary, Zin Thein Kyaw's blog shows us how to use Edge Impulse to easily deploy anomaly detectors in temperature control applications through an example of the application of embedded machine learning in cold chain monitoring. At the same time, ESP32-DevKitC's powerful computing power, compact design and high cost performance also provide ideas and platforms for building innovative applications and simplifying system complexity.