Predict manufacturing machines behaviour to favour preventive maintainance

In order to stay competitive on the market, manufacturing companies needs to maintain high level of efficacy and convenience during the production process. An unforeseen machine breakdown can cause the company high time and financial losses.

Problem

If analytical data are gathered during production cycles, it is possible to identify signs of a future breakdown. The major hindrance in failures predictions is represented by the amount ( generated every year by the maufunacturing sector) and data complexity:

• Multidimensionality: control data include a wide variety of metrics, such as such as feed rate, tool positions, tool loads, and spindle speeds

• Repeated discrete periods of complicated activity: each individual part production cycle will typically follow a very complicated, erratic-looking path through the space of control data variables.

Solutions:

Production cycles control data gathered from manufacturing machines through MtConnect have been preprocessed and utilized for time series analysis and for building classification models for predicting future failures. The following performances have been obtained for future states predictions.