PREDICTIVE MAINTENANCE AND OPTIMIZATION: INUSE INTEGRATES MACHINE LEARNING FUNCTIONS

A new dimension for data analysis

There are different levels of analysis, from the simplest to the most complex.

  • Conditional analysis, which makes it possible to escalate alerts according to predefined conditions upstream. It is found in particular through time series.
  • Predictive analysis, based on simple extrapolations. It makes it possible to evaluate a future value.
  • Frequency-based spectral analysis, used in particular for the IU PREDICT for motors application.

Beyond these different levels, there is the notion of Machine Learning, which takes us one step further in the predictive field.

Focus on Machine Learning: integral part of artificial intelligence

Whereas AI is a “set of theories and techniques used to create machines capable of simulating human intelligence”, Machine Learning algorithms learn autonomously to perform a task or make predictions from data.

The added value lies in the fact that algorithms are learners, that is, they improve their performance over time.

Machine Learning

Full integration of existing Machine Learning models

InUse platform has the ability to integrate and host most machine learning estimators. Those developed internally by the InUse team as well as the external estimators your teams developed in complete autonomy.

The estimators remain your property and the platform provides the computational capabilities necessary for their proper functioning.

The different estimators are treated at the same level as their studio counterparts.

Deployment technical specifications

  • Estimators must be packaged as Python or ONNX pickle files.
  • You then transfer the file to InUse (mandatory phase) which will check for security reasons. 
  • The estimator is integrated into an MLflow instance managed by InUse. This open-source software allows the storage of ML models and deployment.
  • The estimator is finally deployed on a Kubernetes platform allowing it to process a large volume of data without affecting existing models.

Do you wish to have machine learning estimators?

Do you have the in-house skills to develop them? 

We encourage you to submit your estimators within the InUse platform.

Related News