I-care Data Analysis

This project handles data collected on the Intelligent Plant Operations (IPO) demonstrator equipped with I-Care vibration sensors.

http://192.168.10.191:1880/ui/#!/3

The data has been collected in two measurement campaigns, and resulted in two separate datasets.

The sensors data are collected by I-Care's WiCare devices, and then fed to an OSIPi server's ModBus connector, and then eventually loaded into Maximo Monitor.

The sensors channels are:

• channel 1: bearing 1 horizontal (X1)
• channel 2: bearing 1 vertical (Y1)
• channel 3: bearing 2 horizontal (X2)

Labelled condition data

The ConditionData\ folder contains condition data to be used for supervised learning. There are 4 condition classes, with corresponding data files held in subfolders per condition:

• normal: no specific condition
• SI : Structural Imbalance
• WI: Wheel Imbalance
• SIandWI: Structural and Wheel Imbalance There are a collection of JSON files in each of the subfolders, representing the values of the fftv and fftg, rpm, temperature attributes.

In order to use this raw data as a supervised learning training set to submit to AutoAI, it will need to be realigned in a flat .csv file structure.

This is achieved using the code in AutoAI/build_vib_ML_dataset.py, which outputs a ConditionData.csv file out of the contents of the folder.

A second data set had been collected in the form of flat csv files, running the bench in different conditions (no_problem, structural_imbalance, wheel_imbalance and Anomaly). Those files are available in SPSSModel\ folder.

The notebooks\Vibration_EDA_Merger.ipynb notebook is used to display the vibration data under various angles, and collate a merged dataset.