Finding “Variability”
In any batch manufacturing, understanding the variability is one of the fundamental task performed by Process Engineers.
Variability is the deviation of individual parameter (univariate) or a group of parameters (multi-variate) impacting quality, productivity, cost etc. between different batches of the same SKU produced overtime.
Many times, the production process needs to follow a standard pattern ( SOP ) for producing certain SKUs. This standard pattern needs to be followed every time a SKU is being produced in the production line.
Modern Industrial IoT Platforms (IIoT) are equipped with necessary software and hardware interfaces to extract this process information at individual batch level and store the same in their respective databases in near real time. IIoT platforms are also capable of provisioning this contextual information to 3rd party systems and applications to run analytics and derive required insights from it.
In the example below, let us understand how can we leverage simple statistical models in near real-time to detect batch variability leveraging the data on Industrial IoT platforms.
Consider in a batch manufacturing process, maintaining the correct temperature profile is extremely important for the quality and cycle time of the SKU being processed. The LHS graphs shows the temperature profile for 13 batches of the same SKU that was processed in last 3 days. As an operator and line in-charge it is very important for them to monitor and control the temperature profile. Even though everything is done correct due to several natural variations, the temperature profile may not be the exact clone of the SOPs. However, if the operators or the line managers could understand what is the deviation in an “easy to digest “ manner, it can help them to take necessary preventive actions to reduce the deviation and/or variability in subsequent batches.
If you look at the RHS figure, the stat model indicates there are variation wrt the reference temperature profile across the batches produced as indicated by the size of the circles. The x-axis indicates the Batch Serial number and the Size of the circles indicates how much that batch has deviated from the reference temperature profile. Bigger the circle, more is the deviation. If you compare the LHS and RHS and say fig.1 is the reference profile that we want the operators to keep following, clearly maximum variation ( as indicated by the size of the circle ) is exhibited by Batch 13. Now visually if you compare fig.1 and Fig.13 on the LHS you could see the temperature profiles are are not matching at all.
This simple model coupled with the real time data on IIoT platform can provide operation team with insights which they can further use to make informed decision and plan preventive actions.
Y-axis provides an interesting insight, we will discuss in next blog. I will post the code in my GitHub Repo soon.
[ The views expressed in this blog is author’s own views and it does not necessarily reflect the views of his employer, A&M]