Pattern recognition and machine learning for event detection in sensor data


The trend for digital machine data availability in clouds sets in. A new level of monitoring and prediction can be reached. Recorded time-series will be searched for relevant patterns and the life-cycle of machines becomes transparent.

Central of the streamwise approach is a digital analysis of system time series for

  • manufacturing & production events,
  • operation & quality events and
  • maintenance & service events.


The secret of machine life is hidden in the time series data of their operations. We measure an infinite number of sensors already today and usually do not save and analyze these valuable data. Main reason is that humans cannot oversee the vast amount of information, but algorithm do.

streamwise has developed an approach to strategically search in massive system time series data for relevant event patterns and signatures.

In the shown case we measure pressure time series over weeks and longer, identify signatures like the 10 seconds wavelet fluctuation, which shows a single movement. We than train an autoregressive search algorithms to identify that movement or pressure fluctuation wavelet and show or count the occurrence of the movement.

We design predictive, descriptive and decisive models based on advanced data analytics, as well as tailored algorithm solutions for digital monitoring and diagnosis systems. Finally we validate, prototype and build your digital monitoring solution with cloud solutions from our partners network.