The following image shows an arrangement for the Logger/Predictor. An RPI inside will communicate via a 3 core cable also supplying power to an ESP32 in an oil filled box. A lazy loop from the oil filled box will connect to the RPI and another will connect to a protecting housing for the strain gauge. A cloth will hang from the strain gauge. The strain gauge will measure the instantaneous weight of the cloth and send that to the RPi on the inside. The RPi will log weather data and weight of the cloth. As a cross reference some weights will be measured on e.g. a kitchen scales to calibrate the measurements for the cloth.
It is intended to send weights from a digital scales to a RPi SBC that will be able to log data from WUnderground into a database. An output from an ESP32 will allow backup data gathering if the SBC digital output fails.The digital scales will log weights of a patch of cloth that will then be representative of weights of laundry that have been subjected to the same environment. Gathering data from Wunderground that coincides with the weight data will allow a model to be generated linking environmental conditions to drying rates.
The RPi pins 39 GND and 40 GPIO 21 will be used to communicate with the ESP32 via 1-wire protocol.
To enable the one-wire interface you need to add the following line to /boot/config.txt, before rebooting your Pi:
dtoverlay=w1-gpio,gpiopin=21. See this link for more details.
The ESP32 will run communication via 1-wire using the OneWireSlave library it's worth noting that this doesn't seem to support ESP32 hardware by default. For learning about 1-wire this is an interesting site: http://playground.arduino.cc/Learning/OneWire. As a starter I decided to buy a few DS18B20 sensors and interface them to a Raspberry pi. The following page looked like a good starting point for this interfacing : https://www.circuitbasics.com/raspberry-pi-ds18b20-temperature-sensor-tutorial/
The DS18B20 is a temperature sensor but will be useful for proving the concept of interfacing via 1-wire protocol. If this is ever to become something that approaches a product it may be better it were designed as a completely wireless device.
For neural nets I have the following topic to read: https://machinelearningmastery.com/time-series-prediction-lstm-recurrent-neural-networks-python-keras/
