ENBIS: European Network for Business and Industrial Statistics

Prevention is better than cure: machine learning for predictive maintenance

Pierre Harouimi, MathWorks, France

Big Data involves big decisions in smart factories. Connected and automated machines produce real-time data streams, which artificial intelligence algorithms transform into actionable knowledge. Producing this knowledge is often time-consuming with batch processing that takes hours or happens overnight. This mismatch between data and decision creates inefficiencies in the supply chain and maintenance.

The complexity of a supply chain or manufacturing process makes it difficult to manually develop the precise models required for knowledge creation. Machine learning algorithms build these models automatically. Additionally, a new class of machine learning algorithms increases responsiveness and accuracy by dynamically updating their models in near real-time. These “incremental” or “online” learning algorithms process the incoming data into knowledge, then feed this knowledge back into the model: fewer copies of the data, elimination of the time and cost of redistributing the model, ability to manage sets data that exceeds system memory ...

In this 1-hour webinar, you will see:
- Complete workflow: from preprocessing to models deployment
- Incremental machine learning to update streaming data
- MATLAB in a full IT ecosystem, with Kafka, IoTHub Azure & InfluxDB

Short Bio

Pierre Harouimi is an application engineer at MathWorks for 3 years, specialized for Data Analytics & Finance areas. From preprocessing to algorithms deployment, including machine & deep learning models implementation, he covers the whole process of a typical project.

Before joining MathWorks, he was a financial engineer at La Française Asset Management, responsible for the asset allocation & portfolio optimization models.