/abnormality_prediction

Abnormality prediction for industrial air compressors

Primary LanguageJupyter Notebook

Abnormality prediction for industrial air compressors

This repo is for 4th Special 2023 Research and Development Zone AI SPARK Challenge - Air compressor abnormality.

In industrial air compressors and rotating machines, motor and core temperature, vibration, and noise are factors that affect device fatigue, and increased fatigue can cause equipment failure.

The goal is to develop a model that predicts signs of malfunction in industrial equipment through unsupervised learning, so that if fatigue increases, we would prevent equipment failure, and prevents accidents.

We used autoencoder model for this challenge, and secured 104th position on the leaderboard with a macro-F1 score of 0.9538864792.

About Data

train_data : This is training data, which represents all normal cases.
test_data : This is data for evaluation and includes both normal and abnormal cases, and is subject to prediction.
answer_sample : Label file form for submission to be filled out for test_data.

  1. Features
    air_inflow: air intake flow rate (^3/min)
    air_end_temp: Air end temperature (°C)
    out_pressure: Discharge pressure (Mpa)
    motor_current: motor current (A)
    motor_rpm: Motor rotation speed (rpm)
    motor_temp: motor temperature (°C)
    motor_vibe: Motor vibration (mm/s)
    type: Facility number

  2. Additionally, each facility has the following characteristics.
    Facility number 0, 4, 5, 6, 7: 30 HP (horsepower)
    Facility number 1: 20HP
    Facility number 2: 10HP
    Facility number 3: 50HP