Using RapidMiner in Industry 4 Setting?

basisbasis Member Posts: 5 Contributor II
edited November 2018 in Help

Hello All,

New to Forum and I wonder why I didn't use this before as I have been using RapidMiner for educational purposes since 2011. 

Anyways, I am in need of use cases if RapidMiner for big data/ Industry 4.0 setting .Smart factories and such. I know there is a sample process describing predictive maintenance but I wonder whther it is possible to analyze stream data with RapidMiner (for pattern recognition, fault diagnosis, pattern recognition etc.) and if so what is the process like? Also if sample data sets exists that would be great for test. 

Best,

B

Answers

  • IngoRMIngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder

    Hi,

     

    We have actually quite a lot of customer engagements in the manufacturing industry - in fact, this is one of the three largest industries we serve.  You can find a few case studies here: https://rapidminer.com/resources/?type=case-studies

     

    It typically comes down to scenarios like the following:

     

    • Predictive maintenance (for avoiding unplanned outages)
    • Yield optimization (for improving throughput / acceptable output at same costs)
    • Quality assurance (for avoiding low-quality production settings)
    • Root cause analysis (for finding the reasons for problems)
    • Recipe optimization (for avoiding wet lab testing of new products)
    • among many others

    Sensor data is typically involved but that does not necessarily require stream handling in the modeling phase (only in application).  In 99% of all use cases we see, the models are trained offline on batches of data and then the model is applied on data streams in an optimized architecture. RapidMiner perfectly supports this scenario.

     

    Sorry, I can publicly share more insights into what our clients are doing (not even talking about data or processes of course!) but if this is for a commercial project please feel free to reach out to our sales reps to discuss your situation and we might be able to give you more ideas.

     

    Best,

    Ingo

  • basisbasis Member Posts: 5 Contributor II

    Hello Ingo,

    Thanks for fast reply. Case studies, demos and webinars are all good references. But Predictive Maintenance process built in Studio is a very helpful example and similar ones I am looking for. 

    Can you direct me to similar examples for:

    • Yield optimization (for improving throughput / acceptable output at same costs)
    • Quality assurance (for avoiding low-quality production settings)
    • Root cause analysis (for finding the reasons for problems)

    This is mainly for training purposes but can lead to a commercial project as well. 

    Best,

    B

  • IngoRMIngoRM Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, Community Manager, RMResearcher, Member, University Professor Posts: 1,751 RM Founder

    Hi,

     

    All the processes and data I have are customer-related and hence I cannot share, sorry :smileysad:

    Maybe somebody else has something readily available...

     

    However, for those three actually only the Root Cause Analysis can be somewhat special (since it really depends a lot on the data and problem types you are tackling).  Yield optimization and quality assurance can typically be mapped on some form of basic classification or regression problems:

     

    • Yield optimization:
      • For example, you could predict the throughput rate of a production process dependending on process parameters, can be combined with optimization constraints so that other outputs (like vibration or temperature etc.) do not leave given ranges. This than maps directly into a multi-label regression problem.
      • You could also predict if the output of a process will be good enough for further processing / selling / ... and stop the process otherwise (if possible, to save money)  or just run it in a more effective way so that the outcome is JUST good enough.  This than again is a classification problem (in the first case: good enough, yes or no?) or a regression problem (in the second case).
    • Quality assurance:
      • Predict the quality of an outcome based on process parameters and sensor data to focus quality assurance on where it is really necessary. Again, a classification or regression task

     

    Hope this helps a bit,

    Ingo

  • MartinLiebigMartinLiebig Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 3,503 RM Data Scientist

    Hi,

    funny i wrote a quick blog article und prespriptive SPC last week. Have a look here:  https://www.linkedin.com/pulse/overcome-spc-prescriptive-big-data-analytics-dr-martin-schmitz?trk=hp-feed-article-title-publish

     

    ~Martin

    - Sr. Director Data Solutions, Altair RapidMiner -
    Dortmund, Germany
  • basisbasis Member Posts: 5 Contributor II

    Hello Ingo and Martin,

     

    These looks extremely useful thanks!

     

    Best

Sign In or Register to comment.