broken image
broken image
broken image

 

 

  • Our Approach
  • About us 
    • Vision and mission
    • Our team
    • Work for us
    • Awards
    • Blog
  • Manufacturing
  • Infrastructure
  • …  
    • Our Approach
    • About us 
      • Vision and mission
      • Our team
      • Work for us
      • Awards
      • Blog
    • Manufacturing
    • Infrastructure
    Contact us
    broken image
    broken image
    broken image

     

     

    • Our Approach
    • About us 
      • Vision and mission
      • Our team
      • Work for us
      • Awards
      • Blog
    • Manufacturing
    • Infrastructure
    • …  
      • Our Approach
      • About us 
        • Vision and mission
        • Our team
        • Work for us
        • Awards
        • Blog
      • Manufacturing
      • Infrastructure
      Contact us
      broken image

      How to be ‘smart’ with data:

      Dynamic Time Warping

      · Assets,Industry,Data Analysis

      For a ‘smart’ follow-up of machines it is important not to rely on 1 source of truth. In reality this means that, in order to properly assess and predict the efficiency of operations and state-of-health of a machine, combinations are essential. The follow-up system has to combine different data streams (different types of data, such as electrical currents, vibrations, pressures…), but also needs to include different methodologies of data crunching. In the present article we’ll shed a bit of light on the concept of Dynamic Time Warping (DTW): a concept that sounds very futuristic, but has a lot of engineering logic behind.

      What is the Dynamic Time Warping we talk about?

      When continuously measuring data like electrical currents, vibration levels, temperatures… from motors, hoists, rollers this results in so-called Time Series of data: consecutive values with a fixed time spacing in between. Imagine having 2 surface buoys in the sea, in front of the coast, one 5m closer to the shore than the other. We could measure the vertical position of these buoys, resulting in 2 time series of height, in m. When a wave, heading for the shore, passes both buoys and we plot both data streams in a graph, we would first observe a hill in the height data of the buoy furthest from shore, and a bit later a very similar hill in the data from the buoy 5 m closer to shore. A DTW algorithm detects the ‘hill’ in the data and calculates the distance (in time) between both, that is scaled to a 0 - 1 range. If both features coincide, a value of 0 is generated. In case there’s no resemblance at all (e.g. 2 buoys in different oceans) a value of 1 would be produced. This algorithm is deployed continuously on the data streams and, based on the application, the specialist determines the width of the time window as well as the scaling factor to be used.

      When would one use such Dynamic time Warping?

      An interesting field of application can be found in installations where many motors assist in the overall operation, like a metal rolling line, or a waffle production line. In a normal case, each motor emits vibrations related to its own operation and the components it’s directly connected to. In some cases however resonance effects start playing: phenomena appearing in one section of the line have an influence multiple steps down the line. Such situations are often very difficult to identify, and subsequently diagnose. Applying Dynamic Time Warping in this case also can be of enormous added value. In practice the high-resolution RMS vibration levels recorded on all individual motors are fed into the DTW algorithm. All signals are fed pair-wise into the algorithm. This code determines the amplitude of the difference between both signals: what is the time-shift between features present in the signals? In general, on mint installations, one finds a very high resemblance in closely linked operations (close to 0), or no resemblance at all for remote or de-coupled components (value close to 1). This value is followed-up continuously. If however values start decreasing from 1 (or increasing from 0): it means that respectively a resemblance is building up (or the synchronization is getting lost). The code is looking at all pairwise combinations of motors. As such, in case we see a decrease from 1 in a certain stream, we immediately know what motors (or parts of the machine) are affected by cross-talk. Experience learns us that this approach allows detecting deviations way before serious structural issues appear and also helps the engineers in better understanding the phenomenon and providing guidelines on how to react.

      broken image

      As such this Dynamic Time Warping represents a very valuable component in a toolbox for automated and intelligent follow-up of critical installations. When one is capable of detecting cross-talk in an early stage, follow-up damage can be avoided. When on the contrary loss of synchronization can be detected in an early stage, this forms an important asset in fighting quality issues and the related production losses. Despite its very futuristic name, many present-day gains can thus be made.

       

       

      Previous
      Meet our Sales Power team!
      Next
      What's in a name?: ''Infrastructure Analytics''
       Return to site
      Cookie Use
      We use cookies to improve browsing experience, security, and data collection. By accepting, you agree to the use of cookies for advertising and analytics. You can change your cookie settings at any time. Learn More
      Accept all
      Settings
      Decline All
      Cookie Settings
      Necessary Cookies
      These cookies enable core functionality such as security, network management, and accessibility. These cookies can’t be switched off.
      Analytics Cookies
      These cookies help us better understand how visitors interact with our website and help us discover errors.
      Preferences Cookies
      These cookies allow the website to remember choices you've made to provide enhanced functionality and personalization.
      Save