1. Challenge

Challenge #1: Predictive Maintenance

2. Description

We are developing a complete platform for anomaly detection which can take data for several source devices and perform analyses to detect anomalies in each of them.

Intelligently predict when equipment will fail before it does

Using AI condition monitoring to detect various anomalies, signs of deterioration, and other equipment performance issues

Solve the unexpected equipment failures with intelligent algorithms that provide maintenance managers with the information they have always dreamed of having

Lowers operational costs, minimizes downtime, & improves overall asset health & performance 

We are also developing a state-of-the-art event-based data processing backend that allows us to collect data from thousands of sensors and perform analyses on them.

3. Source link code

Machine learning repository: LINK

Event streaming repository: LINK

4. Tech stack

What we have used:

  • Collaboration Tools: Notion (Notes/ToDo’s/Kanban Board), Slack (Communication), GitHub (share code)

  • Data Exploration & Machine Learning: Weka 3.8.6, Google Colab Notebooks, Keras/Tensorflow

  • Data Streaming/Analytics: Apache Kafka deployed on Amazon AWS or Microsoft Azure

  • API communication: REST API build on FastAPI for pushing collected data to Data Streaming


What we should use if this project advances: 

  • Data-Collection: MQTT (Lightweight Protocol for M2M Communication & IoT Data) / CoAP (HTTP)

  • IoT-/Data-Platform: Amazon AWS Greenfield IoT / Microsoft Azure IoT 

  • Standalone Timeseries-Database: Cassandra / Timescale / PostgreSQL

5. Target profile

We are targeting the manufacturing sector. That is, our customers could be manufacturers of industrial machines and robots. The customer could also be the users of said devices.

6. Mock-ups

Dashboard with Data