Time series acquisition with information about Load positioning for a frost-free refrigerator

This dataset contains data collected during an experimental activity conducted on a frost-free refrigerator operating in a real-world environment. The refrigerator (Whirlpool W7 831T MX) was installed in a laboratory office with a constant presence of approximately 8–10 people, thus exposing the system to realistic ambient thermal disturbances and door-opening events. Data acquisition was performed using an Arduino Mega 2560, equipped with a dedicated thermocouple module and an SD card for local data storage. Each row in the dataset corresponds to a single acquisition of all variables, sampled at a frequency of 0.1 Hz (one sample every 10 seconds). The data should be interpreted as a multivariate time series describing the temporal evolution of temperatures and operational variables of the refrigerator under study. The refrigerator was used by occupants without any prescribed usage protocol or behavioural constraints; door openings, load placement, and interactions were entirely driven by normal user behaviour. As a consequence, the recorded dynamics reflect uncontrolled, real-world operating conditions rather than a standardised experimental procedure. Data organisation and file naming The dataset is composed of multiple CSV files named according to the convention: XX_YYY.csv XX identifies the experimental trial. Two trials were conducted at different points in time. YYY identifies the sequential file number within the same trial. When files are ordered sequentially (e.g., 01_001, 01_002, 01_003, …), they reconstruct the full time series of trial XX.Each file represents a continuous acquisition segment, but time variables may restart from zero as specified below. Variables description Each CSV file contains the following columns: Relative acquisition time [s]Time elapsed since the beginning of the current file. Each XX_YYY file restarts from 0. Temperature [°C] – shelf 4 (test variable)* Temperature [°C] – shelf 1 (test variable)* Compressor ON time [s] (test variable)* Compressor OFF time [s] (test variable)* Prediction [string] (test variable)* User feedback [string]String indicating the position of the load inside the refrigerator.This value is updated only after a door opening event, when the user manually inputs a new load position. See the dedicated section below for the allowed user-defined strings. Temperature [°C] – shelf 0 Temperature [°C] – shelf 1 Temperature [°C] – shelf 2 Temperature [°C] – shelf 3 Temperature [°C] – shelf 4See the shelf numbering scheme below. Temperature [°C] – refrigerator compartment probe Ambient room temperature [°C]Temperature of the room where the refrigerator is installed. Condenser temperature [°C]Sensor positioned at the midpoint of the rear tube condenser, located on the back of the refrigerator. Consecutive compressor ON duration [s]Updated when the compressor state is 1 (see variable 18).When the compressor is OFF, the value retains the last ON duration. Consecutive compressor OFF duration [s]Updated when the compressor state is 0 (see variable 18).When the compressor is ON, the value retains the last OFF duration. Compressor state [0,1]Binary variable representing the compressor operating state: 1 = ON 0 = OFF Time since last door closure [ms]Restarts from 0 at the beginning of each file.Resets to 0 every time the door is opened.While the door remains open, the value stays at 0. Technical note on test variables (*) Variables marked with (*) are technical, derived variables introduced during a preliminary validation phase of the acquisition architecture. These variables were used exclusively to verify the computational feasibility of running a proof-of-concept machine learning model directly on the same microcontroller used for data acquisition, under realistic constraints of memory, latency, and integration. They do not represent direct measurements of the observed physical phenomenon, nor are they designed as meaningful features for analytical or predictive applications. The variables are retained in the dataset solely for documentation and traceability of the hardware and pipeline validation process.They are not intended for use in subsequent analyses or in future machine learning models. Their interpretation is strictly dependent on a specific hardware configuration and on an experimental model that is no longer supported or documented; consequently, they are not generalizable nor reproducible in other contexts. The inclusion of these variables is a pragmatic design choice aimed at preserving experimental traceability. Users should be aware that treating them as informative features would introduce methodological errors. Description of possible user input strings In the dataset, the variable User feedback can assume the following values: TOPLOAD: load placed on shelf 4. CENTRALLOAD: load placed on shelves 2–3. BOTTOMLOAD: load placed on shelves 0–1. COOLING: door opening without inserting any load. COOLING2: removal of a load from the refrigerator. SHOPPING: insertion of loads in multiple positions. -1: no label inserted by the user. It is typical for each file to begin with a sequence of -1 values (no string assigned), interrupted by a string and then followed again by -1 values with respect to the “user feedback” variable. This first string is used to confirm that the communication is functioning correctly. Labels may be followed by additional user-defined annotations that were not foreseen by the original test design (e.g., TOPLOAD1).Such extended labels should be considered equivalent to their base label (in this example, TOPLOAD). Shelf numbering refers to the scheme below. REFRIGERATOR SCHEME FOR SHELF NUMBERING -------------------------------------------- | TC4 |-------------------------------------------- | TC3 |-------------------------------------------- | TC2 |-------------------------------------------- | TC1 |-------------------------------------------- | TC0 |---------------------------------------------------------------------------------------- | FREEZER | | COMPARTMENT |--------------------------------------------