Intever Public Database for Arcing Event Detection (IAED)

InteverArcingDataSet-release1.0

-----------------------------------------------------------------

The organization of this database is as follows:

******directory level one******

under the parent folder, there are two sub-folders, [6400_embedded] and [10000_recorder].

Each stands for the Low Resolution data and the High Resolution data.

******directory level two******

Under each folder, there are two sub-folders [multi-load] and [single-load], and each stands for different load configurations.

******directory level three******

Under them we have various data folders, e.g. 'electricoven_m1/', where first part of the file name[electricoven] stands for the appliance name[s] in load configuration, and the second part [m1] or [m2] or [m3] stands for the connection mode.

******directory level four******

Under this final data folder, we have csv files contains voltage and current wav and time_format[1][2].txt file to record the time stamps of each annotation (in seconds).

File time_format[1][2].txt

-----------------------------------------------------------------

This file records the time stamp of each event during our data collection experiment.

There two formats of this file:

Format 1:

each column represents: ["start", "load", "arc", "extinction", "end"]

It is corresponded to the states in experiment: “Start——Turn on Appliance——Generate Arc——Stop Arc——End”

Format 2:

each column represents: ["start","load","arc1","extinction1","arc2","extinction2","load_off","end"]

It is corresponded to the states in experiment: “Start——Turn on Appliance——Generate Arc——Stop Arc——Generate Arc——Stop Arc——Turn off appliance——End”

Script GetData.py

-----------------------------------------------------------------

******General Description******

It is prepared for processing one data folder:

(e.g. D:\THY\DataSet\6400_embedded\single-load\microwaveoven_m1)

the script takes in original current and voltage data (UI.txt) and converts it to [n] rows of sample vector which contains [2*window_size] elements. 

In each row first half of the elements is voltage data and the second half is current data.

window_size could be set to 1000 for HR data and 1024 for LR data due to different sampling period time.

n depends on the numer of data points in one (UI.txt) data file.

According to the time annotation file, time.txt, each row ( sample) is labelled.

It is annotated into 1 (arcing) and 0 (normal).

******Usage Description******

The use of GetData.py requires to prepare two parameters, 

path -- data folder, see example above;

window_size -- determins the number of data points in one sample;

******Function Description******

merge_all_sample(self,task="train",is_save=False)

It merges all UI.txt files under one load configuration (one data folder) into one single unit.

parameter settings: task：train or save；

train -- remove abnormal samples

save -- save all samples

is_save -- to save everything in the orignal data

******Example******

path=r"D:\THY\DataSet\6400_embedded\single-load\microwaveoven_m1"

dp=DataProcess(path,1024)

dp.merge_all_sample(task="train",is_save=True)