Failure Prediction Platform Aimed at Supporting All Types of Equipment

Overview

The Business Analysis Center of Osaka Gas is working on developing a cloud service for failure prediction capable of supporting all types of equipment.

Purpose

Osaka Gas operates various large and small facilities in-house and at customer sites and has been working on preventive maintenance for each of them. In recent years, IoT has been increasingly introduced to the facilities and is expected to accumulate data. Against this backdrop, we aim to improve maintainability and reduce costs by building a failure prediction system leveraging the data.

Data processing flow

Concept

Using its unique logic, this system automatically determines the relevant sensors through computation based on sensor data and carries out calculation using a suitable machine learning model in each case.
In the past in practical data analyses related to failure prediction, it has been necessary to manually identify relevant sensors and select the optimal model. This system automates everything, reducing the manhours required of model developers.