Unsupervised learning for running-in detection in hermetic alternative compressors

Description

Data exploration and application of an unsupervised learning method to aid in the detection of the running-in state in hermetic alternative refrigerant compressors.

The KMeans algorithm was used to group test data with samples of new and already operated compressors, searching for magnitudes whose division matched the expected running-in condition. The "Elbow" method was used to define 3 as the number of groups for clustering, and 120 combinations of preprocessing parameters were evaluated, looking for patterns that indicate detection of the phenomenon.

Getting Started

Pre-requisites

• Anaconda
• Git

Installation

1. Clone the repository

   git clone https://github.com/nicolasantero/compressor-breakin-kmeans-clustering.git

2. Create a virtual environment (optional)

   conda create -n name

3. Activate Virtual Environment (optional)

   conda activate name

4. Installing Packages

pip install requirements.txt

(Início)

Running the notebooks

jupyter notebook

License

Distributed under the MIT License. See LICENSE.txt for more information.

(Início)