Smart Incubator using Nordic Thingy:53

Background

The incubator machine is used to incubate chicken eggs into chicks. Stable conditions are needed to increase the chances of hatching chicks. Incubating chicken eggs using an incubator requires a total incubation period of 21-22 days. The temperature of the hatching room during the incubation period of chicken eggs (18 days) is set at 37°-38°C, while during the hatching period (about 19-21 days) the temperature can be slightly increased to 39°C or stored at 38°C. The humidity is relative, during the incubation period the humidity is maintained at 50% - 55%, and during the hatching period or on the 19-21 day the humidity slightly increases, which is around 60%-65%. In many incubation machines, heaters and blowers are used to obtain stable conditions of temperature and humidity. Generally, these heaters and blowers are turned on continuously so that they consume a lot of energy, when in fact they can be optimized by temporarily turning them off when they have reached a stable temperature and humidity condition by utilizing a control system. It is hoped that the system that will be made can reduce energy consumption used by heaters and blowers but still get optimal conditions for the egg-hatching process.

Solution

I will develop a smart incubator machine that will monitor the temperature and humidity in the incubator. The system that will be made is used to make a decision on whether the heater and blower need to be turned on based on the temperature and humidity conditions that exist in the environment. This system will be built with machine learning for the decision-making process. With this optimization, the heater and blower do not need to be continuously turned on so that less energy is consumed.

How it works

I will use the Nordic Thingy:53 which is equipped with temperature and humidity sensors to carry out the monitoring process in the hatchery incubator. The data obtained will then be processed using machine learning to see if there are unstable temperature and humidity conditions on the incubator machine. When conditions are unstable, the system will take action to change the temperature variable by increasing the temperature on the heating machine or changing the humidity degree variable by activating the fan to improve circulation conditions in the incubator machine. When the machine has reached a stable condition, the heater and blower will be in a state of off/rest for a while so there is no need to consume electricity.

1. Update Firmware (Optional)

Nordic Thingy:53 is installed with the default firmware to connect to the edge impulse, but firmware changes can be made using the nRF Programmer application on GooglePlay Store. Many firmware options are offered, some of which are Edge Impulse, Peripheral LBS, nRF Machine Learning, Matter weather station, etc.

The step that must be done is to just select the firmware that will be used, click install, select the existing Nordic Thingy:53 device, then wait for the upload process to complete. The step that must be done is to just choose the firmware that will be used, click install, select the existing Nordic Thingy:53 device, then wait for the upload process to complete. You can check the image below for more detailed information.

2. Connect to Edge Impulse

The next part is to connect the Nordic Thingy:53 with the edge impulse platform. The thing that must be ensured is that the application is connected to the user's edge impulse account with care login or signup. After the application on the android is connected to the edge impulse account, the next process is to connect the Nordic Thingy:53 with the application via Bluetooth. This process is quite easy because the application will scan automatically to find available devices.

3. Data Acquisition

The next process is the data acquisition process. At this stage, the data collection process is carried out for the incubator machine with the temperature set below 35°C for data with the label "Unoptimal" and conditions above 37°C with the optimal label. Data retrieval is carried out in a prototype incubator machine that already has a heater and fan.

The data retrieval process is carried out using the nRF Edge Impulse application which is already connected to Nordic Thingy:53. The data collection process is carried out by pressing the "RECORD NEW DATA" button on the Data tab. The first process is carried out for the "Unoptimal" label with 10 data samples. Then proceed with the "Optimal" label with the same number of data samples.

4. Training Model

The last part is the model training process, in this section the machine learning model is made using the Edge Impulse platform. This section starts from the Impulse Design stage to create a design for machine learning that will be used. In this section, Preprocessing flatten is selected to process raw data. This process aims to extract the pending features that exist in the raw data which represents the pending points contained in it. The type of model that will be used is classification using hard classification.

Then proceed with the data extraction process. In this section, three main features are selected, namely mean, min, and max. These three features are extracted from the entire existing raw data.

The last step is the model training process using hard classification. In this process, the model training process is carried out with existing data. The accuracy results obtained are 50% accuracy.

Closing

This project is still far from perfect and there is still a lot of development that can be done in it. The developer hopes that in the future there will be more progress so that this project can be made into a project with a better version than the current one.

SEE YOU ON THE NEXT PROJECT! :)