Design Like a Pro: Machine Learning Basics

Travis Cox, Kathy Applebaum, and Kevin McClusky from Inductive Automation will discuss key concepts and best practices, show demos, and answer questions from the audience, to help you start integrating ML into your day-to-day processes.

Learn more about:
• Practical ways to use ML in your factory or facility
• What you'll need to get started
• Existing ML tools and platforms
• And more

1. Moderator/Presenter Travis Cox Co-Director of Sales Engineering Inductive Automation

2. Used by Major Companies Worldwide

3. Co-Presenters Kevin McClusky Co-Director of Sales Engineering, Inductive Automation Kathy Applebaum Senior Software Developer, Inductive Automation

4. Machine learning has existed for a while but has recently come to the forefront. Many companies want to use machine learning but aren’t sure what is it or what to use it for. ML has tremendous potential but it won’t give you instant results. You must answer the right questions and plan your project carefully. Make sure you have the right data, people, and processes. If done right, a ML project can save huge amounts of money and greatly increase your speed after only a few months of development. “What should we do about machine learning?”

5. What We’ll Cover Today We’d like to help by discussing: What machine learning is Ways to apply it Steps to getting started with machine learning Plus: Demo of Azure ML Studio Answering some of your questions

6. What is Machine Learning? Three main branches: Analytics Machine Learning Artificial Intelligence

7. What is Machine Learning? Three main branches: Analytics – Knowledge discovery Descriptive Diagnostic Predictive Prescriptive

8. What is Machine Learning? Three main branches: Analytics – Knowledge discovery Machine Learning – Learn and improve from experience

9. What is Machine Learning? Three main branches: Analytics – Knowledge discovery Machine Learning – Learn and improve from experience Artificial Intelligence – Tasks that simulate human intelligence

10. Software Options for Machine Learning

11. Machine Learning Demo in Ignition

12. What is Machine Learning? Two main types of ML models: Classifiers – predict a category Regression – predict a value

13. Machine Learning Applications #1 application: Predictive Analytics / Predictive Maintenance Examples: Predicting when a motor will fail Predicting when a delivery truck will break down

14. Machine Learning Applications In addition to predictive analytics/maintenance, there are many other industrial ML applications, including: Predicting machine settings Quality control Demand forecasting Raw-material price forecasting Training industrial robots

15. Steps to Machine Learning You need data – lots and lots of quality data ! Collect data from a variety of sources: historical, ERP, etc. Automated data collection, not manual Quality of data more important than quantity

16. Steps to Machine Learning You also need a dedicated person with statistics knowledge and domain knowledge who can label data as good or bad.

17. Steps to Machine Learning Pick a question to answer. Start with what you really want to know Cost function

18. Steps to Machine Learning Use domain knowledge. What data might answer your question? Can you acquire missing data? What quality is your data? Eliminate dependent variables

19. Steps to Machine Learning ETL (Extract, Transform, Load) Can you automate each step? Can new data be acquired automatically? How much clean-up is needed? How will missing values be handled?

20. Steps to Machine Learning Visualize your data. Things to look for: Problem data Obvious trends An obvious algorithm

21. Steps to Machine Learning Determine which algorithm to use. Narrow it down by asking: Classification vs. regression? Labeled vs. unlabeled? Tolerant of missing data? Lazy learning? Retraining? Black box vs. human readable Computing resources available Tolerant of outliers?

22. Steps to Machine Learning Determine which platform to use. Things to think about: Computing resources on-site or in the cloud? How flexible is it? How easy to get data into it? How easy to get results into a usable form? Can it be automated?

23. Steps to Machine Learning Test your model. Testing techniques How accurate? Back up a few steps when needed

24. Software Options for Machine Learning More Control More Automated

25. Azure ML Studio Demo

26. Ignition & Machine Learning Ignition can be a great asset in an effective ML solution: All the data you need to start Connects to ML platforms Ignition & Cirrus Link MQTT Transmission Module connect to AWS Greengrass for machine learning Cirrus Link Cloud Modules enable easy connection of Ignition tag data into AWS or Azure Easier access to machine learning & analytics algorithms coming in Ignition v7.9.8

27. Machine learning is related to analytics and AI Two main ML models are Classifiers (predict category) and Regression (predict value) There are many industrial ML applications; the #1 application is predictive analytics/maintenance To get started, you need a huge amount of data and a dedicated person who is qualified to sort it Other steps: picking a question to answer, using domain knowledge, extract-transform-load, visualize the data, choose an algorithm, choose a platform, and test the model Ignition can be part of an effective ML solution Recap

28. Design Like A Pro Series inductiveautomation.com/ resources

29. Sept. 17-19, 2018 Early-bird tickets on sale now at: icc.inductiveautomation.com

30. Questions & Comments Jim Meisler x227 Vannessa Garcia x231 Vivian Mudge x253 Account Executives: Myron Hoertling x224 Shane Miller x218 Ramin Rofagha x251 Maria Chinappi x264 Lester Ares x214 Kristen Azure x260 Director of Sales: Melanie Hottman 800-266-7798 x247 Co-Directors of Sales Engineering: Travis Cox - x229 [email protected] Kevin McClusky - x237 [email protected]