NexGen Factories- IoT+ML
Today manufacturing industry is facing tremendous challenge on how to structure unorganized data coming from thousands of sensors, vending points, electronic equipment, monitoring machines etc.
Since ‘Data is the new Oil’ , Data scientists are constantly extracting, innovating and transforming tons of raw data into structured meaningful formats for efficient business analysis.
I would like to give some ideas on how the next generation manufacturing plants are taking up technological innovations to double their business growth. IoT, Machine Learning and BlockChain are there 3 pillars that will change the industrial revolution in upcoming years.
Consider my business case,where i am collecting sensors data from boilers in a milk processing plant called Amul (being the largest milk producing company of India). for example you can find hundreds of raw data from https://data.gov.in or Kaggle. This milk processing plant contains various boilers and heating vessels which are constantly running 24x7 to remove bacteria from milk, pasteurization, adding flavors , mixing and cooling etc. This entire workflow is very temperature intensive and sensitive. Any anomaly or deviation in temperature can compromise the end product. Say for example a chocolate ice-cream coated with butter milk and peanut requires the mixture to be heated only between 34.C to 38.C and cooled between -6.C to -8.C. Any such deviation in temperature fluctuations due to equipment overheating or machinery failures can cause the chocolate mixture to compromise its quality. Thus if situations like this are overlooked in the future, companies can face a revenue loss by sending inferior quality items in the market. Similarly sensitive milk items that are consumed by toddlers also need special attention and transparency as milk products coming out of the factories need consistent quality check and verification.
Coming to the technical aspects, lets talk on how i am leveraging Machine Learning, Industrial Internet Of Things and BlockChain to support my business case.
Step 1-The Journey
Here we are collecting raw industrial sensor data, converting it into a JSON and sending it from a Node.js app at an interval of 30s to Azure IoT hub. It is just like temperature sensors in boilers of the plant sending data to IoT hub.
Step 2- Query ,Transform and Alert
Once data reached the IoT hub we will do some real time query on the sensor’s data like SELECT * FROM IOThub where TEMP >40. Once we start receiving high values like 50 we will send the filtered data via Logic apps to the mailbox as an alert OR we can go further by send some Text message alert. “Please monitor the system, the temperature reached 50”..
Step 2- Analysis and Predict
So all these data are constantly pouring into a database say blob or CosmosDB where two ML models are running to predict the Anomaly and machine failures based on historical run. Any anomalous temperature on higher range like 60.C can harm/worn out the system.Using Predictive maintenance once can now have an idea on when the machine will fail. Based on the results we can automated and turn down the boilers from beforehand. Thus saving a lot of bucks on industrial maintenance.
Step 3- Visualize
All these real time data streaming can be visualized in a PowerBI connected dashboard where business people can draw insights on day to day basis without much getting into deeper technical roots.
